WO2021046829A1 - 定位方法、装置及系统 - Google Patents

定位方法、装置及系统 Download PDF

Info

Publication number
WO2021046829A1
WO2021046829A1 PCT/CN2019/105810 CN2019105810W WO2021046829A1 WO 2021046829 A1 WO2021046829 A1 WO 2021046829A1 CN 2019105810 W CN2019105810 W CN 2019105810W WO 2021046829 A1 WO2021046829 A1 WO 2021046829A1
Authority
WO
WIPO (PCT)
Prior art keywords
geometric
geometric features
pose
vehicle
point cloud
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2019/105810
Other languages
English (en)
French (fr)
Inventor
苏斌
袁维平
吴祖光
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huawei Technologies Co Ltd
Original Assignee
Huawei Technologies Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Huawei Technologies Co Ltd filed Critical Huawei Technologies Co Ltd
Priority to PCT/CN2019/105810 priority Critical patent/WO2021046829A1/zh
Priority to EP19930161.5A priority patent/EP3819673A4/en
Priority to CN201980055540.0A priority patent/CN112639882B/zh
Publication of WO2021046829A1 publication Critical patent/WO2021046829A1/zh
Priority to US17/689,642 priority patent/US20220198706A1/en
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
    • G06T7/74—Determining position or orientation of objects or cameras using feature-based methods involving reference images or patches
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
    • G01C21/26—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network
    • G01C21/34—Route searching; Route guidance
    • G01C21/36—Input/output arrangements for on-board computers
    • G01C21/3602—Input other than that of destination using image analysis, e.g. detection of road signs, lanes, buildings, real preceding vehicles using a camera
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/277—Analysis of motion involving stochastic approaches, e.g. using Kalman filters
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W60/00—Drive control systems specially adapted for autonomous road vehicles
    • B60W60/001—Planning or execution of driving tasks
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0231—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means
    • G05D1/0238—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using obstacle or wall sensors
    • G05D1/024—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using obstacle or wall sensors in combination with a laser
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2200/00—Indexing scheme for image data processing or generation, in general
    • G06T2200/04—Indexing scheme for image data processing or generation, in general involving 3D image data
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence
    • G06T2207/10021—Stereoscopic video; Stereoscopic image sequence
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10024—Color image
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10028—Range image; Depth image; 3D point clouds
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30248—Vehicle exterior or interior
    • G06T2207/30252—Vehicle exterior; Vicinity of vehicle

Definitions

  • the invention relates to the technical field of automatic driving in the field of artificial intelligence, and in particular to a positioning method, device and system.
  • AI Artificial Intelligence
  • Autonomous driving is a mainstream application in the field of artificial intelligence.
  • Autonomous driving technology relies on the collaboration of computer vision, radar, monitoring devices, and global positioning systems to allow motor vehicles to achieve autonomous driving without the need for human active operations.
  • Self-driving vehicles use various computing systems to help transport passengers from one location to another. Some autonomous vehicles may require some initial input or continuous input from an operator (such as a navigator, driver, or passenger). The self-driving vehicle permits the operator to switch from the manual operation mode to the automatic driving mode or a mode in between. Since autonomous driving technology does not require humans to drive motor vehicles, it can theoretically effectively avoid human driving errors, reduce the occurrence of traffic accidents, and improve highway transportation efficiency. Therefore, autonomous driving technology is getting more and more attention.
  • the size of point cloud data per kilometer is about 4GB. If the point cloud data is used for positioning, on the one hand, it is necessary to store, transmit and load a large amount of point cloud data to the vehicle’s computer system; on the other hand, it needs to run for real-time vehicle positioning.
  • the registration algorithm is based on a large amount of point cloud data. The calculation takes a long time and it is difficult to meet the real-time requirements of the vehicle for positioning. Especially when the vehicle is in a high-speed motion scene, the real-time positioning of the vehicle poses a huge challenge.
  • the technical problem to be solved by the embodiments of the present invention is to provide a positioning method, device and system to solve the technical problems of large calculation amount in the vehicle positioning process and poor positioning real-time performance.
  • an embodiment of the present invention provides a positioning method, including: a positioning device acquires first point cloud data collected by a vehicle through a point cloud acquisition device; extracting N first geometric features from the first point cloud data, N is a positive integer; further, the first posture of the vehicle is adjusted according to the N first geometric features to obtain the second posture of the vehicle, and the accuracy of the second posture is higher than that of the first posture.
  • first pose and the second pose are both poses obtained by positioning the vehicle, and the accuracy of the second pose is greater than the accuracy of the first pose. It can also be said that the first pose is the predicted pose of the vehicle, and the second pose is the actual pose of the vehicle.
  • the positioning device may be the above-mentioned vehicle or a device on the vehicle, or a terminal, such as a mobile phone, a tablet computer, etc., a positioning chip or a positioning device, or a server or a cloud. It should be understood that the terminal or server may be connected to the vehicle in communication to obtain the first point cloud data observed by the vehicle.
  • the above method provides a positioning method.
  • the first pose with low accuracy is corrected by N first geometric features extracted from the first point cloud data collected by the point cloud acquisition device, and the accuracy can be obtained.
  • the second high pose uses geometric features with a small amount of data for positioning, which greatly reduces the amount of data calculations, and makes vehicle positioning less time-consuming and positioning The real-time performance is good.
  • the positioning device adjusts the first pose of the vehicle according to the N first geometric characteristics to obtain the second pose of the vehicle.
  • An implementation manner may be: the positioning device adjusts the vehicle's second pose according to N first geometric features and N second geometric features in the geometric feature map adjust the first pose of the vehicle to obtain the second pose, where the geometric feature map is the second point cloud data from the point cloud map In the map formed by the geometric features extracted in, the N second geometric features are geometric features that match the N first geometric features.
  • the positioning device adjusts the first pose of the vehicle according to the first geometric feature and the second geometric feature in the geometric feature map to obtain the first implementation of the second pose
  • the way can be:
  • the positioning device determines the transformation relationship between the geometric features according to the N first geometric features and the N second geometric features in the geometric feature map; further, according to the transformation relationship between the geometric features, the first pose of the vehicle is adjusted to Get the second pose.
  • the first pose with low accuracy is corrected by the transformation relationship between the observed first geometric feature and the second geometric feature in the geometric feature map, and a second pose with high accuracy can be obtained.
  • the embodiment of the present application uses geometric features with a small amount of data for registration and positioning, which greatly reduces the amount of data calculations, so that vehicle positioning is less time-consuming and has good real-time positioning.
  • the positioning device determining the transformation relationship between geometric features according to the N first geometric features and the N second geometric features in the geometric feature map may include, but is not limited to, the following two implementation manners
  • the positioning device transforms the N first geometric features by the first transformation amount to obtain N third geometric features, and the third geometric features correspond to the first geometric features one-to-one; further, according to the N third geometric features and the N third geometric features
  • the first error between the second geometric features adjusts the first transformation amount; and, when the number of iterations of the first transformation amount satisfies the stop iteration condition or the first error satisfies the stop iteration condition, the positioning device obtains the first target transformation amount, A target transformation amount is the first transformation amount when the iterative stop condition is satisfied, and the first target transformation amount is used to indicate the transformation relationship between the N first geometric features and the N second geometric features.
  • the solution of the transformation relationship between the geometric features is converted into the minimization of the first error between the N third geometric features and the N second geometric features, and further, iteratively determines that the first error is the smallest when the first error is the smallest.
  • the transformation amount is the transformation relationship between the N first geometric features and the N second geometric features, so that the obtained transformation relationship is more accurate.
  • the positioning device may determine the first error according to the first objective function, where the first objective function may be:
  • a first error ⁇ conversion amount comprises a first rotation R and translation t
  • w i is the weight of the first geometric feature weights V i
  • U i for the second geometric feature of the first geometric feature corresponding to V i
  • the index of the first geometric feature in a geometric feature, i is a positive integer, i ⁇ N.
  • the positioning device transforms the N second geometric features through the second transformation amount to obtain N fourth geometric features, and the fourth geometric features correspond to the second geometric features one-to-one; further, according to the N fourth geometric features and the N fourth geometric features
  • the second error between the first geometric features adjusts the second transformation amount; and, when the number of iterations of the second transformation amount stops the iteration condition or the second error satisfies the stop iteration condition, the positioning device obtains the second target transformation amount, and the second
  • the target transformation amount is the inverse matrix of the second transformation amount when the iterative stop condition is satisfied, and the second target transformation amount is used to indicate the transformation relationship between the N first geometric features and the N second geometric features.
  • the solution of the transformation relationship between the geometric features is converted into the minimization of the second error between the N fourth geometric features and the N first geometric features, and further, iteratively determines the second error when the second error is the smallest.
  • the transformation amount is the transformation relationship between the N first geometric features and the N second geometric features, so that the obtained transformation relationship is more accurate.
  • the positioning device may determine the second error according to the second objective function, where the second objective function may be:
  • second transform includes a rotation amount of R 'and translation t';
  • w i is the weight of the first geometric feature weights V i;
  • u i is the first geometric The vector in the second geometric feature U i corresponding to the feature Vi ;
  • N is the number of the first geometric feature, i is the index of the first geometric feature in the N first geometric features, i is a positive integer, and i ⁇ N.
  • the weight of the first geometric feature weights w i V i first geometric object belongs, wherein V i negative correlation with respect to the distance of the vehicle. That is, the first geometric feature corresponding to the object closer to the vehicle contributes more to the transformation relationship. Due to the first geometric feature closer to the vehicle, the error of the extracted first geometric feature is smaller, so that the positioning device is in Determining the transformation relationship allows the first geometric feature based more on accuracy to improve the accuracy of the transformation relationship, thereby improving the accuracy of positioning.
  • the positioning device adjusts the first pose of the vehicle according to the first geometric feature and the second geometric feature in the geometric feature map to obtain the second implementation of the second pose
  • the way can be:
  • the positioning device estimates the pose of the vehicle according to the first pose to obtain multiple sets of estimated poses; furthermore, determines multiple sets of estimated poses based on N first geometric features and N second geometric features in the geometric feature map Therefore, the positioning device determines the second pose of the vehicle according to the score of each set of estimated poses in the multiple sets of estimated poses, wherein the score of the first set of estimated poses is used to indicate the first set of estimated poses and The closeness of the second pose, the first set of estimated poses is any one of the estimated poses among multiple sets of estimated poses.
  • first pose, the estimated pose, and the second pose are all poses obtained by positioning the vehicle, and the accuracy of the second pose is greater than the accuracy of the first pose. It can also be said that the first pose is the predicted pose of the vehicle, and the second pose is the actual pose of the vehicle.
  • the multiple sets of estimated poses are the poses distributed around the first pose, and the estimated poses whose scores meet the conditions, such as the estimated pose with the largest score is the second pose.
  • the above method provides a vehicle positioning method.
  • the estimated pose is scored by the observed first geometric feature and the second geometric feature in the geometric feature map, and the estimated pose with the highest score is determined as the actual pose of the vehicle.
  • the embodiment of the present application uses geometric features with a small amount of data to perform positioning, which greatly reduces the amount of data calculations, so that vehicle positioning is less time-consuming and has good real-time positioning.
  • the positioning device determines multiple sets of estimated pose scores according to the N first geometric features and the N second geometric features in the geometric feature map, which may include but not limited to the following three implementations :
  • the positioning device determines the estimated value of the first parameter corresponding to each set of estimated pose according to each set of estimated pose and N second geometric features; determines the first parameter according to the first pose and N first geometric characteristics Further, the positioning device determines the score of each group of estimated poses according to the error between the estimated value of the first parameter corresponding to each set of estimated poses and the observed value of the first parameter.
  • the first parameter is at least one of a distance, an azimuth angle, and an altitude angle; the estimated value of the first parameter corresponding to each set of estimated poses is the N second geometric features relative to each set of estimated poses.
  • the first parameter of the vehicle in the first pose; the observed value of the first parameter is the first parameter of the N first geometric features respectively relative to the vehicle in the first pose.
  • the above method converts the estimated pose score into an error between the estimated value of the first parameter under the estimated pose and the actual observation value of the first parameter, so that the estimated pose evaluation process is simpler, and the positioning time is further reduced.
  • the positioning device separately transforms the N second geometric features through the transformation relationship between each set of estimated poses and the first poses, and the N fifth geometric features corresponding to each set of estimated poses are obtained.
  • the second geometric features There is a one-to-one correspondence with the fifth geometric features; further, the positioning device determines the score of each set of estimated poses according to the errors between the N fifth geometric features and the N first geometric features corresponding to each set of estimated poses.
  • the above method converts the estimated pose score into the error between the estimated value of the first parameter under the estimated pose and the actual observation value, so that the estimated pose evaluation process is simpler, and the positioning time is further reduced.
  • the positioning device separately transforms the N first geometric features through the transformation relationship between each set of estimated poses and the first poses, and the N sixth geometric features corresponding to each set of estimated poses are obtained.
  • the first geometric features There is a one-to-one correspondence with the sixth geometric features; further, the positioning device determines the score of each set of estimated poses according to the errors between the N sixth geometric features and the N second geometric features corresponding to each set of estimated poses.
  • the positioning device before the positioning device adjusts the first pose of the vehicle according to the N first geometric features to obtain the second pose of the vehicle, the positioning device may also obtain the first pose of the vehicle .
  • the positioning device to obtain the first pose of the vehicle can include but is not limited to the following two implementations:
  • the positioning device determines the predicted pose of the vehicle at the current moment according to the second pose at the previous moment.
  • the predicted pose at the current moment is the first pose of the vehicle acquired by the positioning device, and the previous moment is the current moment.
  • the positioning device can input the accurate pose of the vehicle at the last moment (that is, the second pose at the last moment) and the control parameters of the vehicle at the last moment into the kinematics equation of the vehicle to predict the first position at the current moment. Posture.
  • the above method estimates the pose at the previous moment with high accuracy, and the estimated first pose at the current moment is closer to the actual pose of the vehicle, reducing the number of iterations of the calculation, and further improving the second pose of the vehicle.
  • the calculation efficiency and response speed of vehicle positioning is the following:
  • Implementation (2) The positioning device determines the first position of the vehicle according to the positioning system and determines the first posture of the vehicle according to the inertial sensor.
  • the first posture includes the first position and the first posture.
  • the positioning device adjusts the first pose of the vehicle according to the first geometric feature and the second geometric feature in the geometric feature map to obtain a third implementation of the second pose
  • the way can be:
  • the positioning device determines the predicted pose of the vehicle at the current moment according to the second pose at the previous moment and the control parameters of the vehicle at the previous moment, and the previous moment is the moment before the current moment; further, the positioning device passes the second parameter
  • the error between the observed value of and the predicted value of the second parameter updates the predicted pose of the vehicle to obtain the second pose of the vehicle, where the observed value of the second parameter is based on the first observed value of the vehicle in the first pose
  • a geometric feature is determined, and the predicted value of the second parameter is determined based on the predicted pose and the second geometric feature in the geometric feature map.
  • the positioning device determines the first position of the vehicle according to the positioning system and according to the inertial sensor.
  • the first posture of the vehicle is determined, and the first posture includes a first position and a first posture.
  • first pose, the predicted pose at the current moment, and the second pose at the current moment are all poses obtained by positioning the vehicle at the current moment.
  • the accuracy of the second pose is greater than the accuracy of the predicted pose, and it is also greater than the accuracy of the first pose.
  • first pose and the predicted pose at the current moment are the predicted poses of the vehicle at the current moment, and the second pose is the actual pose of the vehicle at the current moment.
  • the predicted pose at the current moment is calculated based on the error between the observed value of the second parameter determined by the observed first geometric feature and the predicted value of the second parameter determined based on the second geometric feature in the geometric feature map. Update to obtain the actual pose of the vehicle.
  • the embodiment of the present application uses geometric features with less data for positioning, which greatly reduces the amount of data calculations and makes vehicle positioning less time-consuming , Real-time positioning is good.
  • the positioning device determines the difference between the observed value of the second parameter and the predicted value of the second parameter based on the first pose, the predicted pose, the N first geometric features, and the N second geometric features.
  • One way to realize the error between the two can be: the positioning device determines the predicted value of the second parameter based on the predicted pose and N second geometric features; determines the observation of the second parameter based on the first pose and N first geometric features Value; further, the error between the observed value of the second parameter and the predicted value of the second parameter is determined according to the observed value of the second parameter and the predicted value of the second parameter.
  • the above method adopts the Kalman filtering method to determine the second pose, with less update times, reducing the calculation process, and fast positioning.
  • the second parameter is at least one of a distance, an azimuth angle, and an altitude angle;
  • the predicted value of the second parameter is the second parameter of the N second geometric features respectively relative to the vehicle in the predicted pose;
  • second The observed value of the parameter is the second parameter of the N first geometric features respectively relative to the vehicle in the first pose.
  • the above method measures the error between the predicted pose and the actual pose of the vehicle by predicting the error between the predicted value of the second parameter and the actual observed value of the second parameter, and updates the predicted pose so that the second parameter
  • the error between the predicted value of and the actual observation value of the second parameter is the smallest, that is, the actual pose of the vehicle, that is, the second pose is obtained.
  • the above further reduces the time-consuming positioning.
  • the positioning device can be the above-mentioned vehicle or a device on the vehicle, or a terminal, such as a mobile phone, a tablet computer, etc., or a positioning device, a positioning chip, a server or a cloud, etc.
  • the terminal or The server may communicate with the vehicle to obtain the first geometric feature observed by the vehicle.
  • the first point cloud data is the information of the points on the surface of the object observed by the vehicle represented in the space determined by the first pose, and N first geometric features Each first geometric feature in is used to indicate the geometric feature of an object observed by the vehicle in the space determined by the first pose.
  • the positioning device extracts the geometric features of the object from the point cloud data with a large amount of data, that is, the N first geometric features, and performs registration and positioning through the geometric features with a small amount of data, which greatly reduces the amount of data calculations. , Improve the response speed of positioning.
  • the method further includes: the positioning device searches the geometric feature map for N second geometric features that match the N first geometric features.
  • the positioning device searching for N second geometric features matching the N first geometric features in the geometric feature map may include an implementation manner: Find N second geometric features that match the N first geometric features in the first area of the map.
  • the first area is an area determined based on the first pose, and the first area is not smaller than the point cloud collection device of the vehicle. Scan range.
  • the positioning device determines the first area according to the first pose, reduces the search range, and improves calculation efficiency.
  • the positioning device searching for N second geometric features matching the N first geometric features from the first area of the geometric feature map may include, but is not limited to, the following two implementation manners:
  • the above method performs matching by calculating the deviation between two geometric features to improve the matching accuracy.
  • the positioning device may be implemented in a manner based on the determined second geometric features and geometric characteristics of the first V i matches a plurality of geometric features U i is phase matched with the above-described first geometric feature genus deviation V i of the first geometric feature V i smallest geometrical feature.
  • an implementation manner for the positioning device to extract the N first geometric features from the first point cloud data may be: the positioning device recognizes N objects in the first point cloud data; and further, The positioning device determines the first geometric feature of each object based on the point cloud data of each of the N objects.
  • the first object is any one of a plurality of objects
  • the first geometric feature of the first object is determined based on the point cloud data of the first object as an example to illustrate the point cloud based on each of the N objects
  • the data determines the first geometric feature of each object:
  • the point cloud data of the first object is fitted with a straight line to obtain the first geometric characteristic of the first object, and the first geometric characteristic of the first object is the geometric characteristic of the fitted straight line ;
  • the geometric shape of the first object is a curve
  • the geometric shape of the first object is a plane
  • the geometric shape of the first object is a curved surface
  • surface fitting is performed on the point cloud data of the first object to obtain the first geometric characteristic of the first object, and the first geometric characteristic of the first object is the geometric characteristic of the curve obtained by the curved surface.
  • the above method provides a method for extracting the first geometric feature from the point cloud data. First identify the geometric shape of the object, and then perform the fitting through the fitting method corresponding to the identified geometric shape, and use the geometric features of the line/curve/plane/surface obtained by fitting as the first geometric feature of the object, which can improve the extraction The accuracy of the first geometric feature.
  • an embodiment of the present application also provides a geometric feature extraction method, including: an execution device acquires point cloud data to be processed; extracting at least one geometric feature from the point cloud data to be processed; wherein the at least one Geometric features are used for vehicle positioning.
  • the execution device may be a geometric feature map generating device, and the point cloud data to be processed may be the second point cloud data in the point cloud map.
  • the point cloud data to be processed may be the second point cloud data in the point cloud map.
  • the extracted geometric features form a geometric feature map, and the geometric feature map is used for positioning the vehicle through the geometric features.
  • the execution device may be a positioning device
  • the point cloud data to be processed may also be the first point cloud data collected by the vehicle through the point cloud acquisition device.
  • the N first geometric features extracted from the data are used to adjust the first pose of the vehicle to obtain the second pose of the vehicle.
  • an implementation manner for the execution device to extract at least one geometric feature from the to-be-processed point cloud data may be:
  • the execution device recognizes at least one object in the cloud data to be processed; further, the positioning device determines the geometric feature of each object based on the point cloud data of each object in the at least one object.
  • the second object is any one of the at least one object, and the geometric feature of the second object is determined based on the point cloud data of the second object as an example to illustrate the determination of each object based on the point cloud data of each object in the at least one object.
  • the geometric characteristics of each object, and the point cloud data of the second object can determine the geometric characteristics of the second object by:
  • the execution device performs a straight line fitting on the point cloud data of the second object to obtain the geometric feature of the second object, and the geometric feature of the second object is the geometric feature of the straight line obtained by fitting;
  • the execution device performs curve fitting on the point cloud data of the second object to obtain the geometric characteristic of the second object, and the geometric characteristic of the second object is the geometric characteristic of the curve obtained by fitting;
  • the execution device performs plane fitting on the point cloud data of the second object to obtain the geometric feature of the second object, and the geometric feature of the second object is the geometric feature of the plane obtained by fitting;
  • the execution device performs curved surface fitting on the point cloud data of the second object to obtain the geometric characteristic of the second object, and the geometric characteristic of the second object is the geometric characteristic of the curve obtained by the curved surface.
  • an embodiment of the present application also provides a positioning device, including:
  • the first obtaining unit is used for the first point cloud data collected by the vehicle through the point cloud collecting device;
  • the feature extraction unit is used to extract N first geometric features from the first point cloud data, where N is a positive integer;
  • the adjustment unit is configured to adjust the first pose of the vehicle according to the N first geometric characteristics to obtain a second pose of the vehicle, and the accuracy of the second pose is higher than that of the first pose degree.
  • the device further includes:
  • the second acquisition unit is configured to acquire the first pose of the vehicle before the adjustment unit adjusts the first pose of the vehicle according to the N first geometric characteristics to obtain the second pose of the vehicle.
  • the device further includes:
  • the matching unit is used to search for N second geometric features matching the N first geometric features in the geometric feature map.
  • the positioning device of the third aspect described above also includes other units used to implement the positioning method described in the first aspect.
  • each unit of the positioning device or other units please refer to the relevant information in the first aspect. Description, I won’t repeat it here.
  • an embodiment of the present application also provides a positioning device, including: a processor and a memory, the memory is used to store a program, the processor executes the program stored in the memory, and when the program stored in the memory is executed, The method according to the first aspect or any one of the first aspect of the claim can be implemented.
  • the above positioning device also includes other devices or modules used to implement the positioning method described in the first aspect.
  • each device or other device of the above positioning device please refer to the relevant description in the above first aspect, which will not be repeated here. .
  • an embodiment of the present application also provides a geometric feature extraction device, including:
  • the obtaining unit is used to obtain the point cloud data to be processed
  • the extraction unit is configured to extract at least one geometric feature from the point cloud data to be processed; wherein, the at least one geometric feature is used for vehicle positioning.
  • the geometric feature extraction device may be a geometric feature map generating device, and the point cloud data to be processed may be the second point cloud data in the point cloud map.
  • the geometric features extracted from the cloud data form a geometric feature map, and the geometric feature map is used to locate the vehicle through the geometric features.
  • the geometric feature extraction device may be a positioning device, and the point cloud data to be processed may also be the first point cloud data collected by the vehicle through the point cloud acquisition device.
  • the N first geometric features extracted from the cloud data are used to adjust the first pose of the vehicle to obtain the second pose of the vehicle.
  • geometric feature extraction device described in the fifth aspect above also includes other units used to implement the geometric feature extraction method described in the second aspect, and the specific implementation of each unit of the above positioning device or other units Please refer to the related description in the above second aspect, which will not be repeated here.
  • an embodiment of the present application also provides a geometric feature extraction device, including: a processor and a memory, the memory is used to store a program, the processor executes the program stored in the memory, when the program stored in the memory When executed, the method described in the second aspect or any one of the second aspects can be implemented.
  • the above positioning device also includes other devices or modules used to implement the geometric feature extraction method described in the second aspect.
  • each device or other device of the above positioning device please refer to the relevant description in the above second aspect. No longer.
  • an embodiment of the present application also provides a vehicle, including: a point cloud collection device, a processor, and a memory, the processor is connected to the point cloud collection device through a bus, and the point cloud collection device is used to collect point clouds Data, the memory is used to store a program, the processor executes the program stored in the memory, and when the program stored in the memory is executed, the method as described in the first aspect or any one of the first aspects can be implemented.
  • an embodiment of the present application also provides a vehicle, including: a point cloud collection device, a processor, and a memory, the processor is connected to the point cloud collection device through a bus, and the point cloud collection device is used to collect point clouds Data, the memory is used to store a program, the processor executes the program stored in the memory, and when the program stored in the memory is executed, the method as described in the second aspect or any one of the second aspect can be implemented.
  • an embodiment of the present application also provides a computer-readable storage medium, the computer storage medium stores a computer program, the computer program includes program instructions, and when executed by a processor, the program instructions can implement the first The method in the aspect.
  • an embodiment of the present application also provides a computer-readable storage medium, the computer storage medium stores a computer program, the computer program includes program instructions, and when executed by a processor, the program instructions can implement the second The method in the aspect.
  • the embodiments of the present application also provide a computer program product containing instructions, which when the computer program product runs on a computer, causes the computer to execute the method in the first aspect.
  • the embodiments of the present application also provide a computer program product containing instructions, which when the computer program product runs on a computer, causes the computer to execute the method in the above second aspect.
  • a positioning chip in a thirteenth aspect, includes a processor and a data interface.
  • the processor reads instructions stored in a memory through the data interface and executes the method in the first aspect.
  • the chip may further include a memory in which instructions are stored, and the processor is configured to execute the instructions stored in the memory, and when the instructions are executed, the processor is configured to execute the first The method in the aspect.
  • a chip in a fourteenth aspect, includes a processor and a data interface.
  • the processor reads instructions stored on a memory through the data interface and executes the method in the second aspect.
  • the chip may further include a memory in which instructions are stored, and the processor is configured to execute the instructions stored in the memory, and when the instructions are executed, the processor is configured to execute the second The method in the aspect.
  • an electronic device which includes the positioning device in any one of the above-mentioned third aspect or the fourth aspect.
  • an electronic device which includes the positioning device in any one of the above-mentioned fifth aspect or the sixth aspect.
  • a positioning method may include: a positioning device receives N first geometric features, and further, the positioning device adjusts the first pose of the vehicle according to the N first geometric features to obtain the vehicle's The second pose, where the N first geometric features are extracted from the first point cloud data, the first point cloud data is the point cloud data collected by the vehicle through the point cloud acquisition device, and the accuracy of the second pose Higher than the accuracy of the first pose.
  • the specific implementation of the positioning device adjusting the first pose of the vehicle according to the N first geometric features to obtain the second pose of the vehicle may refer to the relevant description in the first aspect above, and will not be repeated here.
  • the way for the positioning device to obtain the N first geometric features may be: the positioning device receives the N first geometric features sent by the terminal/vehicle, and the N first geometric features are collected by the terminal/vehicle from the vehicle through the point cloud Extracted from the first point cloud data collected by the device.
  • the specific implementation of the terminal/vehicle extracting the N first geometric features from the first point cloud data collected by the vehicle through the point cloud acquisition device can refer to the positioning device from the first point cloud data in the first aspect mentioned above.
  • the specific implementation of extracting the N first geometric features will not be repeated here.
  • the positioning device adjusts the first posture of the vehicle according to the N first geometric features to obtain the second posture of the vehicle.
  • the relevant description in the first aspect above please refer to the relevant description in the first aspect above, which will not be repeated here. .
  • an embodiment of the present application also provides a positioning device, including:
  • the receiving unit is configured to receive N first geometric features, where the N first geometric features are extracted from first point cloud data, and the first point cloud data are points collected by the vehicle through the point cloud acquisition device Cloud data
  • the adjustment unit is configured to adjust the first pose of the vehicle according to the N first geometric characteristics to obtain a second pose of the vehicle, and the accuracy of the second pose is higher than that of the first pose degree.
  • the device further includes:
  • the acquiring unit is configured to acquire the first pose of the vehicle before the adjustment unit adjusts the first pose of the vehicle according to the N first geometric characteristics to obtain the second pose of the vehicle.
  • the device further includes:
  • the matching unit is used to search for N second geometric features matching the N first geometric features in the geometric feature map.
  • the positioning device described in the eighteenth aspect above further includes other units for implementing the positioning method described in the seventeenth aspect.
  • the specific implementation of each unit of the positioning device or other units please refer to the above paragraph. The description of the seventeen aspects will not be repeated here.
  • an embodiment of the present application also provides a positioning device, including: a processor and a memory, the memory is used to store a program, the processor executes the program stored in the memory, and when the program stored in the memory is executed ,
  • a positioning device including: a processor and a memory, the memory is used to store a program, the processor executes the program stored in the memory, and when the program stored in the memory is executed ,
  • the method described in the seventeenth aspect or any one of the seventeenth aspects can be implemented.
  • the above positioning device also includes other devices or modules for implementing the positioning method described in the seventeenth aspect.
  • each device or other device of the above positioning device please refer to the related description in the above seventeenth aspect. Go into details again.
  • the embodiments of the present application also provide a computer-readable storage medium, the computer storage medium stores a computer program, the computer program includes program instructions, and the program instructions, when executed by a processor, can realize The method in the seventeen aspects.
  • the embodiments of the present application also provide a computer program product containing instructions, which when the computer program product runs on a computer, causes the computer to execute the method in the seventeenth aspect.
  • a positioning chip in a twenty-second aspect, the chip includes a processor and a data interface, the processor reads instructions stored in a memory through the data interface, and executes the method in the seventeenth aspect.
  • the chip may also include a memory in which instructions are stored, and the processor is configured to execute the instructions stored in the memory, and when the instruction is executed, the processor is configured to execute the tenth The method in the seven areas.
  • an electronic device which includes the positioning device in any one of the eighteenth aspect or the nineteenth aspect described above.
  • a positioning method may include: first point cloud data collected by a vehicle through a point cloud acquisition device; further, the vehicle extracts N first geometric features from the first point cloud data Acquire N first geometric features, and then, the vehicle sends the N first geometric features to the positioning device, so that after the positioning device receives the N first geometric features, the positioning device will The feature adjusts the first pose of the vehicle to obtain the second pose of the vehicle, wherein the accuracy of the second pose is higher than the accuracy of the first pose.
  • the specific implementation of the positioning device adjusting the first pose of the vehicle according to the N first geometric features to obtain the second pose of the vehicle may refer to the related description in the first aspect above, and will not be repeated here.
  • implementation of the vehicle extracting N first geometric features from the first point cloud data can refer to the specific implementation of the positioning device extracting N first geometric features from the first point cloud data in the first aspect. I won't repeat them here.
  • an embodiment of the present application also provides a positioning device, including:
  • the collection unit is used for the first point cloud data collected by the point cloud collection device
  • An extraction unit configured to extract N first geometric features from the first point cloud data to obtain N first geometric features
  • the sending unit is configured to send the N first geometric characteristics to the positioning device, so that after the positioning device receives the N first geometric characteristics, the positioning device will perform the first position of the vehicle according to the N first geometric characteristics.
  • the pose is adjusted to obtain the second pose of the vehicle, wherein the accuracy of the second pose is higher than the accuracy of the first pose.
  • the device further includes:
  • the acquiring unit is configured to acquire the first pose of the vehicle before the adjustment unit adjusts the first pose of the vehicle according to the N first geometric characteristics to obtain the second pose of the vehicle.
  • the device further includes:
  • the matching unit is used to search for N second geometric features matching the N first geometric features in the geometric feature map.
  • the positioning device described in the above-mentioned twenty-fifth aspect further includes other units for implementing the positioning method described in the twenty-fourth aspect.
  • the specific implementation of each unit of the above-mentioned positioning device or other units please refer to The relevant description in the above twenty-fourth aspect will not be repeated here.
  • an embodiment of the present application also provides a positioning device, including a processor, a memory, and a communication interface.
  • the memory is used to store a program.
  • the processor executes the program stored in the memory. When the program is executed, the method described in any one of the twenty-fourth aspect or the twenty-fourth aspect can be implemented.
  • the aforementioned positioning device also includes other devices or modules for implementing the positioning method described in the twenty-fourth aspect.
  • each device or other device of the aforementioned positioning device please refer to the relevant description in the aforementioned twenty-fourth aspect. I won't repeat it here.
  • the embodiments of the present application also provide a computer-readable storage medium, the computer storage medium stores a computer program, the computer program includes program instructions, and the program instructions, when executed by a processor, can realize such as The method in the twenty-fourth aspect.
  • the embodiments of the present application also provide a computer program product containing instructions, which when the computer program product runs on a computer, causes the computer to execute the method in the twenty-fourth aspect.
  • a chip in a twenty-ninth aspect, includes a processor and a data interface.
  • the processor reads instructions stored on a memory through the data interface and executes the method in the twenty-fourth aspect.
  • the chip may further include a memory in which instructions are stored, and the processor is configured to execute the instructions stored in the memory, and when the instructions are executed, the processor is configured to execute the second Fourteen methods.
  • an electronic device which includes the positioning device in any one of the above-mentioned twenty-fifth aspect or the twenty-sixth aspect.
  • an embodiment of the present application also provides a vehicle, including: a point cloud acquisition device, a processor, and a memory, the processor is connected to the point cloud acquisition device through a bus, and the point cloud acquisition device is used to collect Point cloud data, the memory is used to store a program, the processor executes the program stored in the memory, and when the program stored in the memory is executed, any implementation such as the twenty-fourth aspect or the twenty-fourth aspect can be realized The method described.
  • FIG. 1A is a schematic explanatory diagram of a road scene provided by an embodiment of the present invention.
  • FIG. 1B is a schematic explanatory diagram of a geometric feature corresponding to FIG. 1A provided by an embodiment of the present invention
  • FIG. 2 is a schematic diagram of the framework of a system provided by an embodiment of the present invention.
  • FIG. 3 is a functional block diagram of a vehicle provided by an embodiment of the present invention.
  • FIG. 4A is a schematic flowchart of a geometric feature extraction method provided by an embodiment of the present invention.
  • 4B is a schematic diagram of the principle of extracting geometric features of road surfaces and curbs according to an embodiment of the present invention.
  • 4C is a schematic diagram of another principle for extracting geometric features of road surfaces and curbs according to an embodiment of the present invention.
  • FIG. 4D is a schematic diagram of the principle of extracting the geometric characteristics of the geometric characteristics of the road indication line according to an embodiment of the present invention.
  • 4E is a schematic diagram of distribution of point cloud data of trees according to an embodiment of the present invention.
  • 4F is a schematic explanatory diagram of the geometric features of a billboard provided by an embodiment of the present invention.
  • 4G is a schematic explanatory diagram of the geometric features of a street light pole provided by an embodiment of the present invention.
  • 4H is a schematic explanatory diagram of the distance and angle of a point on an indicator provided by an embodiment of the present invention.
  • FIG. 4I is a schematic diagram of the distribution of point cloud data of a building provided by an embodiment of the present invention.
  • FIG. 5 is a schematic flowchart of a method for generating a geometric map according to an embodiment of the present invention
  • FIG. 6 is a schematic diagram of a positioning principle according to an embodiment of the present invention.
  • FIG. 7 is a schematic flowchart of a positioning method provided by an embodiment of the present application.
  • FIG. 8A is a schematic explanatory diagram of a first area provided by an embodiment of the present application.
  • FIG. 8B is a schematic explanatory diagram of another first area provided by an embodiment of the present application.
  • FIG. 8C is a schematic explanatory diagram of still another first area provided by an embodiment of the present application.
  • FIG. 9A is a schematic flowchart of another positioning method provided by an embodiment of the present application.
  • FIG. 9B is a schematic flowchart of a first positioning method provided by an embodiment of the present application.
  • FIG. 9C is a schematic flowchart of a second positioning method provided by an embodiment of the present application.
  • FIG. 10A is a schematic flowchart of another positioning method provided by an embodiment of the present application.
  • FIG. 10B is a schematic flowchart of a third positioning method provided by an embodiment of the present application.
  • FIG. 10C is a schematic explanatory diagram of a third positioning manner provided by an embodiment of the present application.
  • 10D is a schematic explanatory diagram of the elevation angle and the azimuth angle of a vector provided by an embodiment of the present application.
  • FIG. 11A is a schematic flowchart of another positioning method provided by an embodiment of the present application.
  • FIG. 11B is a schematic flowchart of a fourth positioning method provided by an embodiment of the present application.
  • FIG. 11C is a schematic flowchart of a fifth positioning method provided by an embodiment of the present application.
  • FIG. 11D is a schematic flowchart of a fifth positioning method provided by an embodiment of the present application.
  • FIG. 12 is a schematic structural diagram of a positioning device provided by an embodiment of the present application.
  • FIG. 13 is a schematic diagram of the hardware structure of a positioning device provided by an embodiment of the present application.
  • FIG. 14 is a schematic structural diagram of a geometric feature extraction device provided by an embodiment of the present application.
  • FIG. 15 is a schematic diagram of the hardware structure of another geometric feature extraction device provided by an embodiment of the present application.
  • point cloud data is a collection of points in a three-dimensional space, which is usually represented by three-dimensional coordinates in the space, and is the point information obtained by transforming the information on the surface of the object.
  • the data of a point can also include RGB color, gray value, depth, intensity information, segmentation results, and so on.
  • the intensity information is the intensity of the echo received by the lidar, which is related to the surface material, roughness, incident angle of the target, the energy and wavelength of the laser radar emitted wave, etc.; the result of the segmentation can be the object's Identification, attributes, location or area of the point on the object, etc.
  • C ⁇ P 1 , P 2 , P 3 ,..., P n ⁇ represents a group of point cloud data.
  • point cloud data is collected by a point cloud collection device such as a laser radar (laser radar), a stereo camera (stereo camera), and a time of flight (TOF) camera.
  • a point cloud collection device such as a laser radar (laser radar), a stereo camera (stereo camera), and a time of flight (TOF) camera.
  • laser radar laser radar
  • stereo camera stereo camera
  • TOF time of flight
  • Vehicles equipped with point cloud collection devices such as lidar sensors and dual cameras can collect the point cloud data of each road. Then, through map construction methods, such as simultaneous localization and map construction (simultaneous localization and mapping, SLAM) algorithms, the points Cloud data is accumulated frame by frame to construct a point cloud map.
  • map construction methods such as simultaneous localization and map construction (simultaneous localization and mapping, SLAM) algorithms
  • the points Cloud data is accumulated frame by frame to construct a point cloud map.
  • the point cloud map is usually point cloud data in the world coordinate system (English: world coordinate system, also called global coordinate system).
  • Point cloud registration is the process of unifying two sets of point cloud data from different coordinate systems (also called point set space) into a specified coordinate system through rigid transformations such as rotation and translation.
  • the two point clouds to be registered can be completely overlapped with each other after transformation such as rotation and translation. Therefore, point cloud registration is to find the coordinate transformation relationship between the two point clouds.
  • the point cloud data collected by the point cloud collection device of the vehicle is in the current inaccurate position (also called “predicted position” in this application) and inaccurate posture (also called in this application). Is the point cloud data in the coordinate system determined by the "predicted posture”).
  • the accurate position also referred to as the "actual position” in this application
  • the accurate posture of the vehicle can be obtained (Also referred to as "actual posture” in this application).
  • the actual position refers to the position of the vehicle in the world coordinate system
  • the actual posture refers to the accurate posture of the vehicle.
  • “pose” in this application refers to "position” and “attitude”.
  • predicted pose refers to predicted position and predicted pose
  • actual pose refers to actual position And actual posture.
  • the coordinate transformation relationship between the above two sets of point cloud data can be calculated through a matching algorithm.
  • Pose refers to the position and posture of the vehicle in a certain coordinate system.
  • position refers to the position coordinates of the vehicle in the coordinate system
  • attitude refers to the rotation angle of the vehicle around the x-axis, y-axis and z-axis in the coordinate system, which are respectively the pitch angle (pitch), yaw angle (yaw),
  • the roll angle (roll) has a total of 6 degrees of freedom (DoF).
  • the pose can also be selected as 3DoF to express, that is, the x-axis coordinate, the y-axis coordinate And the yaw angle. Furthermore, the dimensionality of the point cloud data can be reduced, and the data volume of the point cloud data can be reduced. It should be understood that the expression of the above pose is only an exemplary description, and the pose in this application can also be represented by 4DoF (that is, x-axis coordinates, y-axis coordinates, z-axis coordinates, and yaw angle) or other degrees of freedom. , This application does not limit this.
  • the size of point cloud data per kilometer is about GB or more than ten GB. If point cloud data is used for positioning, on the one hand, it is necessary to store, transmit and load a large amount of point cloud data to the computer system of the vehicle; on the other hand, On the one hand, the registration algorithm that needs to be run for real-time vehicle positioning is based on a large amount of point cloud data. The calculation takes a long time and it is difficult to meet real-time requirements. Especially when the vehicle is in a high-speed motion scene, the real-time performance of the algorithm is huge. challenge.
  • the positioning method proposed in this application is not based on the registration of point cloud data, but based on the geometric features extracted from the point cloud data.
  • Geometric features are extracted from point cloud data, and point cloud data describes information about sampling points on the surface of objects in the environment.
  • the geometric shapes of objects in the environment in the embodiments of the present application mainly include lines and surfaces, where the surfaces may include planes and curved surfaces, and the lines may include straight lines and curves.
  • An object may include, but is not limited to, multiple geometric features, and the object described by a geometric feature may be a part of the object whose geometric shape is a line or a surface.
  • the geometric features of an object whose geometric shape is a plane can include the normal vector of the plane, the position of the plane, etc.; the geometric feature of the surface can include multiple normal vectors, multiple principal directions, multiple coordinate points on the surface, and coefficients of polynomial surfaces, etc.
  • the geometric characteristics of a straight line may include a direction vector and the position of a straight line; the geometric characteristics of a curve may include at least one of multiple tangent vectors, multiple normal vectors, multiple coordinate points, coefficients of a polynomial curve, etc. kind.
  • Each vector in the above geometric feature is the direction vector and/or position of the object described by the geometric feature, and the direction vector can be represented by a vector, where the vector can be a unit vector.
  • the geometric feature of the road surface that is, the normal vector of the plane where the road surface is located, can be expressed as (0,0,1). It should be noted that the above (0,0,1) is only an exemplary description.
  • the direction vector of the roadside is (0.8, 0.6, 0).
  • the position in the above geometric feature is the position of the object described by the geometric feature, which can be represented by coordinates.
  • the object may be identified first, and then the point cloud data of the object may be fitted with a straight line/curve/plane/curved surface to obtain the geometric characteristics of the determined object. It should be understood that in the embodiments of the present application, it is not necessary to obtain the geometric features of all objects in the environment, but to obtain the geometric features of some objects in the environment, for example, perform geometric features of fixed objects or parts on objects in the environment. extract.
  • the object or part of the object (also referred to as the object in the embodiment of the present application) for which geometric features need to be acquired may be roadsides, road surfaces, traffic index lines, traffic poles, traffic signs, Lane lines, tree trunks, tunnels, buildings, etc. It should be understood that in the embodiments of the present application, an object may be referred to as an object, or a part of an object may be referred to as an object.
  • An object corresponds to a geometric feature.
  • a traffic rod may include a cross rod and a vertical rod, where the cross rod may correspond to a linear geometric feature, and the vertical rod may correspond to a linear geometric feature.
  • a rectangular parallelepiped-shaped building may include an upper edge, a lower edge, a left edge, a right edge, a main plane, and a side surface, etc., wherein the upper edge, the lower edge, the left edge, and the right edge may respectively correspond to a linear geometric feature ,
  • the main plane and one side can respectively correspond to a plane geometric feature.
  • a tree trunk can correspond to a linear geometric feature.
  • the attributes of the objects in the point cloud data can also be identified, for example, the classification and size to which the objects belong. This attribute is helpful for more accurate matching of geometric features. The matching of geometric features will be introduced in other parts, so I won't repeat them here.
  • the category of the object can be divided according to the category of the object to which the object belongs, including roadsides, road surfaces, road indicator lines, traffic poles, traffic signs, lane lines, tree trunks, buildings, and so on.
  • the above-mentioned categories can also be further divided.
  • a building includes the plane, left edge, right edge, upper edge, lower edge, etc. of the building;
  • a traffic pole can include a cross bar and a vertical pole, etc.;
  • a road indicator line can include a solid line, a dashed line, and a turning Line etc.
  • the types of objects can also be divided by the geometric shapes of the objects. At this time, the types of objects can include straight line shapes, curved shapes, plane shapes, and curved surfaces. It should be understood that the types of objects It may also include a combination of the above two or classification in other ways, etc., which is not limited here.
  • the size may be the length of the line, the area of the plane, etc., which are not limited in the embodiment of the present application.
  • the size is the category of the object described by the geometric feature, which can correspond to the length, width, height, or surface area of the object.
  • the position may be the position of the object described by the geometric feature. Since the position corresponds to the object one-to-one, the position is also called the position of the geometric feature.
  • the position of the geometric feature can be the position coordinate of the object described in the coordinate system adopted by the point cloud data, and the selection rule of the position coordinate of the geometric feature can be agreed upon.
  • the position coordinates of the geometric feature can be the position coordinates of the geometric center of the object corresponding to the geometric feature; or, the position coordinates of the geometric feature are the position coordinates of the lowest position of the object corresponding to the geometric feature (that is, the z-axis in the point cloud data of the object).
  • the position coordinate corresponding to the point with the smallest coordinate); or, the position coordinate of the object is the coordinate corresponding to the point with the smallest x-axis and/or y-axis coordinate in the point cloud data of the object. It should be noted that in the embodiments of the present application, other standards may also be used to set the position of the object described by the geometric feature, which is not limited.
  • different selection rules may also be adopted for different classified objects to determine the position coordinates of the geometric feature corresponding to the object.
  • the geometric feature is a plane, and the position coordinates of the geometric feature can be the position coordinates of the geometric center of the main plane; for another example, for a tree trunk, the set feature is a straight line, and the position coordinates of the geometric feature It can be the position coordinate corresponding to the point with the smallest z-axis coordinate in the point cloud data of the tree trunk; for another example, for the upper edge of a building, the geometric feature is a straight line, and the position coordinate of the geometric feature can be a point on the upper edge of the building The position coordinate corresponding to the point with the smallest x-axis coordinate in the cloud data.
  • the first method the coordinates of any position in the point cloud data of the object are selected as the position of the geometric feature corresponding to the object. It should be understood that the position at this time can be used to reflect the relative positional relationship between various objects.
  • the second way other objects can be used as a reference, and the position of the geometric feature corresponding to the object can be determined according to the position of the referenced object.
  • its position may be the midpoint of the positions of two objects located on both sides of the road.
  • FIG. 1A The schematic illustration of the road scene shown in FIG. 1A and the schematic illustration of the geometric features shown in FIG. 1B.
  • a, b, c, d, and e are five road scene schematic diagrams.
  • f, g, h, i, and j are the geometric features extracted from the point cloud data of the road scene shown in a, b, c, d, and e respectively.
  • the left edge, right edge, and upper edge of a building can respectively correspond to a linear geometric feature, which can include the direction vector and position of a straight line; the wall of a building can correspond to a plane
  • Geometric features for example, include the normal vector and the position of the geometric center of the plane.
  • the horizontal bar and the vertical bar in the traffic pole can respectively correspond to a linear geometric feature.
  • the edge of a straight line shape can correspond to a straight line geometric feature.
  • the inner wall of the tunnel can correspond to a geometric feature of a curved surface.
  • the multiple walls of the support column may respectively correspond to a plane geometric feature.
  • the edges of the support columns can respectively correspond to a linear geometric feature.
  • the geometric feature map is a map formed by geometric features extracted from the point cloud data of the point cloud map.
  • the geometric feature map may also include the attributes of the geometric feature, such as size, category, etc., and may also include the location and the address corresponding to the location, the road name corresponding to the geometric feature of the road, and the like.
  • the geometric feature map can be used as a layer in a map (such as a Google map, a Baidu map, a Gaode map, or a point cloud map), or can be used as a separate map.
  • the vehicle or the terminal bound to the vehicle can load the geometric feature map, or only the data of the geometric feature map, which is not limited in the embodiment of the present application.
  • the matching of geometric features refers to matching two sets of geometric features in different coordinate systems, and then establishes the correspondence between geometric features in different coordinate systems.
  • the point cloud data collected by the vehicle in real time is the point data in the coordinate system determined based on the predicted pose (that is, the inaccurate position and posture) of the vehicle.
  • the geometric features extracted from the point cloud data match the geometric features in the geometric feature map. It should be understood that the two geometric features that establish the correspondence relationship are essentially the expression of an object in different coordinate systems. Before locating the vehicle, it is first necessary to find a set of geometric features in the geometric feature map that matches a set of geometric features extracted from the point cloud data collected by the vehicle.
  • the matching between the two sets of geometric features can be established according to at least one of the positions, directions, and attributes of the two sets of geometric features.
  • the relevant description in the following positioning method please refer to the relevant description in the following positioning method, which will not be omitted here. Go into details.
  • the 3D geometric features of roadsides, traffic poles, traffic signs, lane lines, tree trunks, buildings and other objects are distributed in a three-dimensional space with non-parallel and non-coplanar features.
  • two or more non-coplanar geometric features can determine the transformation relationship between the two coordinate systems, and obtain the vehicle positioning result. It should be understood that in the positioning process, the more geometric features, the more accurate the positioning result will be.
  • Kalman filter is an optimal state estimation method. Assuming a discrete linear dynamic system, the state of the next moment can be inferred based on the state of the previous moment.
  • the prediction equation can be expressed as Among them, the subscript "k” indicates the current moment, the subscript "k-1" indicates the previous moment, the symbol “ ⁇ ” in the state indicates that the state is an estimated value, and the superscript " ⁇ ” in the state indicates that the state is based on the previous
  • the predicted result of the state Represents the predicted estimated value of the state at the current moment, a k is the control input of the vehicle at the current moment, such as acceleration, steering, etc. All predictions contain noise. The greater the noise, the greater the uncertainty.
  • the state prediction noise is represented by the covariance matrix (covariance), which is usually represented by P.
  • the observed value of the system is z k , according to the prediction result of the system state Can predict the observed value of the system among them, H k is the observation matrix of the system.
  • Optimal state estimate Is the residual error between the observed value and the predicted value of the observation
  • the Kalman coefficient (also known as Kalman gain) K k to update namely Among them, H k is the observation matrix Represents the residual error between actual observations and expected observations. Multiplying the residual error by the coefficient K k can be used to predict the state Make corrections.
  • the Kalman gain matrix K k actually represents the estimated coefficient when the variance P between the system state and the predicted state is the smallest in the state optimal estimation process.
  • the first pose, the second pose, the estimated pose, the predicted pose, etc. are all measured values of the pose of the vehicle, and the accuracy of each pose is different.
  • the accuracy in the embodiments of this application refers to the degree of closeness of the measured pose of the vehicle, such as the first pose, second pose, estimated pose, predicted pose, etc., to the real pose of the vehicle, that is, the accuracy of the first pose
  • the degree indicates the difference between the first posture and the real posture of the vehicle.
  • the accuracy of the second posture refers to the gap between the second posture and the real posture of the vehicle. A low accuracy indicates a large gap, and a high accuracy indicates a small gap.
  • the purpose of the embodiments of the present application is to obtain a second pose with high accuracy by adjusting the first pose with low accuracy. That is to say, compared to the first pose, the second pose is closer to the real pose of the vehicle, and can be used as the real pose of the vehicle (also known as the actual pose). Therefore, the first pose can be called The predicted pose, the second pose is called the actual pose. It should be understood that the second pose is also a measurement of the actual pose of the vehicle, but it is closer to the actual pose of the vehicle than the first pose.
  • an embodiment of the present application provides a system architecture.
  • the system 10 may include: a data collection device 110, a database 120, a point cloud map server 130, a geometric feature map generation device 140, and a map server 150.
  • the vehicle 160, the terminal 180, and the positioning server 190 can all be used as positioning devices.
  • the present invention does not specifically address whether the positioning device is a vehicle, a terminal, or a positioning server. limited.
  • the data collection device 110 is used to collect point cloud data through a point cloud collection device and store the point cloud data in the database 120, and may be a vehicle equipped with a point cloud collection device, or other devices that can realize point cloud data collection. In another implementation, the data collection device 110 may also extract geometric features in the collected point cloud data, and store the geometric features in the data block 120.
  • the point cloud map server 130 is used to form a point cloud map from the point cloud data collected by the data collection device 110, and can also receive a request for point cloud data for a specific area from the positioning device, and respond to the request to send the point cloud data to the positioning device. Point cloud data for a specific area.
  • the geometric feature map generating device 140 may be a device with computing functions such as a server or a computer, and is used to obtain the point cloud data of the point cloud map from the database 120 or the point cloud map server 130, and then extract the geometric features in the point cloud data. Get the geometric feature map. Further, the geometric feature map can be stored in the database 130 or stored locally. The geometric feature map generating device 140 may also receive a request for acquiring the geometric feature of the first region sent by the positioning device, and then send the geometric feature of the first region to the positioning device in response to the request. It should be understood that the generation of the geometric feature map and the function of the responsive positioning device can also be implemented by different devices, which are not limited in the embodiment of the present application.
  • the map server 150 may be a server of a map application, such as a server of a map application such as a server of Baidu Maps, a server of Google Maps, and a server of AutoNavi Maps.
  • the positioning device can establish a communication connection with the map server 150 to perform data interaction, so that the positioning device can navigate to the destination according to the currently positioned pose.
  • the geometric feature map generating device 140 may also include data such as addresses and roads, and the positioning device may also establish a communication connection with the geometric feature map generating device 140 to realize navigation of the positioning device.
  • the positioning device may be a vehicle 160, a computer system on the vehicle 160, or a terminal 180 connected to the vehicle 160, such as a mobile phone, a tablet computer, etc.; it may also be a positioning server 190.
  • the vehicle 160 is equipped with a point cloud collection device, an inertial sensor, and the like.
  • the positioning device is the positioning server 190 or the terminal 180
  • the vehicle 190 may send the collected first point cloud data and the first pose of the current environment to the positioning server 190 or the terminal 180, and the positioning server 190/terminal 180 can according to the received
  • the first point of cloud data and the first position are used to adjust the first pose of the vehicle 160, and send the first pose obtained by positioning to the vehicle 160.
  • the positioning server 190 may be a server that provides positioning requirements for vehicles.
  • the positioning server 190 can obtain the first point cloud data collected by the vehicle 160 through the point cloud collection device, or can pre-store the geometric feature map to realize the positioning of the vehicle; further, the positioning server can also obtain the positioning The second pose of is sent to the vehicle 160 or the terminal 180.
  • the above-mentioned point cloud acquisition device may specifically be at least one laser radar, and the laser radar may be a multi-line radar laser sensor, such as a 4-line, 8-line, 16-line, 32-line, 64-line or other number of line beam laser radars.
  • the point cloud acquisition device may also be a stereo camera, a TOF camera, etc., where the stereo camera may include multiple cameras, or may include a combination of one or more cameras and lidar.
  • Inertial sensors may include, but are not limited to, one or more of gyroscopes, accelerometers, and magnetic sensors.
  • the computer system of the vehicle can send the motion information collected by the inertial sensor and the point cloud data collected by the point cloud collection device, or the geometric features extracted from the point cloud data, to the positioning device, so that the positioning device can receive To locate the data.
  • the positioning device can request the geometric feature map generating device 140 to obtain the geometric feature map, or download the geometric feature map from the geometric feature map generating device 140, and then combine the geometric features and geometric features extracted from the point cloud data collected by the point cloud collection device.
  • the geometric features in the feature map are matched, and further, the geometric features obtained by the matching are used for positioning.
  • the specific positioning method is that the positioning device can determine the first pose of the vehicle, which is an estimate of the current position of the vehicle, with low accuracy; the positioning device obtains the points collected by the point cloud collection device on the vehicle Cloud data, the point cloud data is based on the description in the coordinate system determined by the first pose; further, multiple first geometric features are extracted from the point cloud data, and then the geometric feature maps are searched for and the multiple A plurality of second geometric features that match the first geometric feature, where the first geometric feature and the second geometric feature have a one-to-one correspondence, and further, because the multiple first geometric features are based on inaccurate first pose pairs
  • the multiple second geometric features describe the expression of the object, so the positioning device can obtain the accurate pose of the vehicle, that is, the second pose, according to the multiple first geometric features and the multiple second geometric features.
  • point cloud map server 130 and map server 150 are not necessary components of the system 10 in the embodiment of the present application, and the system 10 may also include other devices, which are not limited here.
  • Fig. 3 is a functional block diagram of a vehicle provided by an embodiment of the present invention.
  • the vehicle is configured in a fully or partially autonomous driving mode.
  • the vehicle 100 can control itself while in the automatic driving mode, and can determine the current state of the vehicle and its surrounding environment through human operations, determine the possible behavior of at least one other vehicle in the surrounding environment, and determine the other vehicle
  • the confidence level corresponding to the possibility of performing the possible behavior is to control the vehicle 100 based on the determined information.
  • the vehicle 100 can be placed to operate without human interaction.
  • the vehicle 100 may be the vehicle 160 in the system shown in FIG. 2 and may include various subsystems, such as a traveling system 102, a sensing system 104, a control system 106, one or more peripheral devices 108, and a power supply 111, a computer system 112, and User interface 116.
  • the vehicle 100 may include more or fewer subsystems, and each subsystem may include multiple elements.
  • each subsystem and element of the vehicle 100 may be interconnected by wire or wirelessly.
  • the travel system 102 may include components that provide power movement for the vehicle 100.
  • the propulsion system 102 may include an engine 118, an energy source 119, a transmission 122, and wheels/tires 121.
  • the engine 118 may be an internal combustion engine, an electric motor, an air compression engine, or a combination of other types of engines, such as a hybrid engine composed of a gas oil engine and an electric motor, or a hybrid engine composed of an internal combustion engine and an air compression engine.
  • the engine 118 converts the energy source 119 into mechanical energy.
  • Examples of energy sources 119 include gasoline, diesel, other petroleum-based fuels, propane, other compressed gas-based fuels, ethanol, solar panels, batteries, and other sources of electricity.
  • the energy source 119 may also provide energy for other systems of the vehicle 100.
  • the transmission device 122 can transmit the mechanical power from the engine 118 to the wheels 121.
  • the transmission 122 may include a gearbox, a differential, and a drive shaft.
  • the transmission device 122 may also include other devices, such as a clutch.
  • the drive shaft may include one or more shafts that can be coupled to one or more wheels 121.
  • the sensing system 104 may include several sensors that sense information about the environment around the vehicle 100.
  • the sensing system 104 may include a global positioning system 122 (the global positioning system 122 may be a GPS system, a Beidou system or other positioning systems), an inertial measurement unit (IMU) 124, and a point cloud acquisition device 126
  • the point cloud acquisition device 126 may include a laser radar 127, a stereo camera 128, a TOF camera, and the like.
  • the sensing system 104 may also include sensors of the internal system of the monitored vehicle 100 (for example, an in-vehicle air quality monitor, a fuel gauge, an oil temperature gauge, etc.). Sensor data from one or more of these sensors can be used to detect objects and their corresponding characteristics (position, shape, direction, speed, etc.). Such detection and identification are key functions for the safe operation of the autonomous vehicle 100.
  • the point cloud collection device 126 can be used to obtain point cloud data of the surrounding environment of the vehicle 100 to achieve accurate estimation of the geographic location and posture of the vehicle 100.
  • the global positioning system 122 in the vehicle 100 such as GPS, BPS, etc., can be used to roughly estimate the geographic location of the vehicle 100.
  • the IMU 124 is used to sense the position and orientation change of the vehicle 100 based on the inertial acceleration, and can be used to roughly estimate the attitude of the vehicle.
  • the IMU 124 may be a combination of an accelerometer and a gyroscope.
  • the point cloud acquisition device 126 may specifically be at least one laser radar 127, and the laser radar 127 may be a multi-line laser radar, such as a 4-line, 8-line, 16-line, 32-line, 64-line, or other number of line beams.
  • the lidar 127 may use radio signals to sense objects in the surrounding environment of the vehicle 100.
  • the point cloud collection device 126 can also be used to sense the speed and/or the forward direction of the object.
  • the point cloud acquisition device 126 may also be a stereo camera 128, where the stereo camera 128 may include multiple cameras, or may include one or more cameras and lidar sensors. The stereo camera may be used to capture multiple images of the surrounding environment of the vehicle 100 including depth information.
  • the control system 106 controls the operation of the vehicle 100 and its components.
  • the control system 106 may include various components, including a steering system 132, a throttle 134, a braking unit 136, a sensor fusion algorithm 138, a computer vision system 141, a route control system 142, and an obstacle avoidance system 144.
  • the steering system 132 is operable to adjust the forward direction of the vehicle 100.
  • it may be a steering wheel system in one embodiment.
  • the throttle 134 is used to control the operating speed of the engine 118 and thereby control the speed of the vehicle 100.
  • the braking unit 136 is used to control the vehicle 100 to decelerate.
  • the braking unit 136 may use friction to slow down the wheels 121.
  • the braking unit 136 may convert the kinetic energy of the wheels 121 into electric current.
  • the braking unit 136 may also take other forms to slow down the rotation speed of the wheels 121 to control the speed of the vehicle 100.
  • the computer vision system 141 may be operable to process and analyze the images captured by the camera 130 in order to identify objects and/or features in the surrounding environment of the vehicle 100.
  • the objects and/or features may include traffic signals, road boundaries, and obstacles.
  • the computer vision system 141 may use object recognition algorithms, Structure from Motion (SFM) algorithms, video tracking, and other computer vision technologies.
  • SFM Structure from Motion
  • the computer vision system 141 may be used to map the environment, track objects, estimate the speed of objects, and so on.
  • the route control system 142 is used to determine the travel route of the vehicle 100.
  • the route control system 142 may combine data from the sensor 138, the GPS 122, and one or more predetermined maps to determine the driving route for the vehicle 100.
  • the obstacle avoidance system 144 is used to identify, evaluate and avoid or otherwise cross over potential obstacles in the environment of the vehicle 100.
  • control system 106 may add or alternatively include components other than those shown and described. Alternatively, a part of the components shown above may be reduced.
  • the vehicle 100 interacts with external sensors, other vehicles, other computer systems, or users through peripheral devices 108.
  • the peripheral device 108 may include a wireless communication device 146, an onboard computer 148, a microphone 151, and/or a speaker 152.
  • the peripheral device 108 provides a means for the user of the vehicle 100 to interact with the user interface 116.
  • the onboard computer 148 may provide information to the user of the vehicle 100.
  • the user interface 116 can also operate the onboard computer 148 to receive user input.
  • the on-board computer 148 can be operated through a touch screen.
  • the peripheral device 108 may provide a means for the vehicle 100 to communicate with other devices located in the vehicle.
  • the microphone 151 may receive audio (eg, voice commands or other audio input) from the user of the vehicle 100.
  • the speaker 152 may output audio to the user of the vehicle 100.
  • the wireless communication device 146 may wirelessly communicate with one or more devices directly or via a communication network.
  • the wireless communication device 146 may use 3G cellular communication, such as CDMA, EVDO, GSM/GPRS, or 4G cellular communication, such as LTE. Or 5G cellular communication.
  • the wireless communication device 146 may use WiFi to communicate with a wireless local area network (WLAN).
  • the wireless communication device 146 may directly communicate with the device using an infrared link, Bluetooth, or ZigBee.
  • the wireless communication device 146 may also include one or more dedicated short-range communications (DSRC) devices, which can implement public and/or private communication between vehicles and/or roadside units (RSU). data communication.
  • DSRC dedicated short-range communications
  • the power source 111 can provide power to various components of the vehicle 100.
  • the power source 111 may be a rechargeable lithium ion or lead-acid battery.
  • One or more battery packs of such batteries may be configured as a power source to provide power to various components of the vehicle 100.
  • the power source 111 and the energy source 119 may be implemented together, such as in some all-electric vehicles.
  • the computer system 112 may include at least one processor 113 that executes instructions 115 stored in a non-transitory computer readable medium such as the data storage device 114.
  • the computer system 112 may also be multiple computing devices that control individual components or subsystems of the vehicle 100 in a distributed manner.
  • the processor 113 may be any conventional processor, such as a commercially available CPU. Alternatively, the processor may be a dedicated device such as an ASIC or other hardware-based processor.
  • FIG. 3 functionally shows the processor, memory, and other elements of the computer 110 in the same block, those of ordinary skill in the art should understand that the processor, computer, or memory may actually include Multiple processors, computers, or memories stored in the same physical enclosure.
  • the memory may be a hard disk drive or other storage medium located in a housing other than the computer 110. Therefore, a reference to a processor or computer will be understood to include a reference to a collection of processors or computers or memories that may or may not operate in parallel. Rather than using a single processor to perform the steps described here, some components such as steering components and deceleration components may each have its own processor that only performs calculations related to component-specific functions .
  • the processor may be located away from the vehicle and wirelessly communicate with the vehicle.
  • some of the processes described herein are executed on a processor arranged in the vehicle and others are executed by a remote processor, including taking the necessary steps to perform a single manipulation.
  • the data storage device 114 may include instructions 115 (eg, program logic), which may be executed by the processor 113 to perform various functions of the vehicle 100, including those functions described above.
  • the data storage device 114 may also contain additional instructions, including sending data to, receiving data from, interacting with, and/or interacting with one or more of the propulsion system 102, the sensing system 104, the control system 106, and the peripheral device 108. The instruction to control.
  • the data storage device 114 can also store data, such as the first point cloud data collected by the point cloud collection device 126, the first geometric feature extracted from the first point cloud data, the geometric feature map, the road map, and the route. Information, the location, direction, speed of the vehicle, and other such vehicle data, as well as other information. Such information may be used by the vehicle 100 and the computer system 112 during the operation of the vehicle 100 in autonomous, semi-autonomous, and/or manual modes.
  • the vehicle 100 or the computer system 112 may determine the first pose of the vehicle based on the positioning system 122, such as the global positioning system 122 and the inertial measurement unit 124, to roughly estimate the current position of the vehicle; and then obtain the point cloud collection on the vehicle
  • the first point cloud data collected by the device 126 further, a plurality of first geometric features are extracted from the first point cloud data, and the first geometric features are in a coordinate system determined based on the first pose.
  • each object search for multiple second geometric features matching the multiple first geometric features from the geometric feature map, where the first geometric feature and the second geometric feature are in one-to-one correspondence, and further, because The multiple first geometric features are based on the expression of the object described by the multiple second geometric features based on the inaccurate first pose. Therefore, the vehicle 100 or the computer system 112 can be based on the multiple first geometric features and the multiple second geometric features. Geometric features, get the exact pose of the vehicle, that is, the second pose.
  • the user interface 116 is used to provide information to or receive information from a user of the vehicle 100.
  • the user interface 116 may include one or more input/output devices in the set of peripheral devices 108, such as a wireless communication device 146, an in-vehicle computer 148, a microphone 151, and a speaker 152.
  • the computer system 112 may control the functions of the vehicle 100 based on inputs received from various subsystems (eg, the travel system 102, the sensing system 104, and the control system 106) and from the user interface 116. For example, the computer system 112 may utilize input from the control system 106 in order to control the steering unit 132 to avoid obstacles detected by the sensing system 104 and the obstacle avoidance system 144. In some embodiments, the computer system 112 is operable to provide control of many aspects of the vehicle 100 and its subsystems.
  • various subsystems eg, the travel system 102, the sensing system 104, and the control system 106
  • the computer system 112 may utilize input from the control system 106 in order to control the steering unit 132 to avoid obstacles detected by the sensing system 104 and the obstacle avoidance system 144.
  • the computer system 112 is operable to provide control of many aspects of the vehicle 100 and its subsystems.
  • one or more of these components described above may be installed or associated with the vehicle 100 separately.
  • the data storage device 114 may exist partially or completely separately from the vehicle 1100.
  • the aforementioned components may be communicatively coupled together in a wired and/or wireless manner.
  • FIG. 3 should not be construed as a limitation to the embodiment of the present invention.
  • An autonomous vehicle traveling on a road can recognize objects in its surrounding environment to determine the adjustment to the current speed.
  • the object may be other vehicles, traffic control equipment, or other types of objects.
  • each recognized object can be considered independently, and based on the respective characteristics of the object, such as its current speed, acceleration, distance from the vehicle, etc., can be used to determine the speed to be adjusted by the self-driving car.
  • the self-driving car vehicle 100 or the computing device associated with the self-driving vehicle 100 may be based on the characteristics of the identified object and the surrounding environment
  • the state of the object e.g., traffic, rain, ice on the road, etc.
  • each recognized object depends on each other's behavior, so all recognized objects can also be considered together to predict the behavior of a single recognized object.
  • the vehicle 100 can adjust its speed based on the predicted behavior of the identified object.
  • an autonomous vehicle can determine what stable state the vehicle will need to adjust to (for example, accelerating, decelerating, or stopping) based on the predicted behavior of the object.
  • other factors may also be considered to determine the speed of the vehicle 100, such as the lateral position of the vehicle 100 on the road on which it is traveling, the curvature of the road, the proximity of static and dynamic objects, and so on.
  • the computing device can also provide instructions to modify the steering angle of the vehicle 100 so that the self-driving car follows a given trajectory and/or maintains an object near the self-driving car (for example, , The safe horizontal and vertical distances of cars in adjacent lanes on the road.
  • the above-mentioned vehicle 100 may be a car, truck, motorcycle, bus, boat, airplane, helicopter, lawn mower, recreational vehicle, playground vehicle, construction equipment, tram, golf cart, train, and trolley, etc.
  • the embodiments of the invention are not particularly limited.
  • the method can be executed by the vehicle 160, the positioning device, the geometric feature map generating device 140 in FIG. 2 or the vehicle 100 in FIG. Take an example to illustrate. Please refer to the geometric feature extraction method shown in FIG. 4A.
  • the method may include but is not limited to the following steps:
  • the point cloud data to be processed is the first point cloud data of the surrounding environment acquired by the vehicle through the point cloud acquisition device.
  • the geometric features extracted from the first point cloud data are multiple first geometric features or N first geometric features in the third embodiment, and the extracted geometric features are used for vehicle positioning.
  • the first point cloud data may be one frame of point cloud data or multiple frames of point cloud data, or point cloud data obtained after inter-frame point cloud superimposition is performed on multiple frames of point cloud data.
  • the embodiments of this application are not limited.
  • the positioning device may establish a communication connection with the vehicle, and then receive the first point cloud data acquired by the vehicle through the point cloud collection device.
  • the point cloud data to be processed is the second point cloud data obtained by the geometric feature map generating device from a database, a data collection device, or a point cloud map.
  • the geometric features extracted from the to-be-processed point cloud data are used to generate a geometric feature map.
  • the point cloud data to be processed obtained by the geometric feature map generating device can be the point cloud data of the area to be processed, or it can be one or more frames of point cloud data to be processed, or it can be multi-frame point cloud data.
  • the point cloud data obtained after the point clouds are superimposed between frames is not limited in this embodiment of the present application.
  • the vehicle can also perform noise reduction processing on the point cloud data to be processed, for example, filter out outliers in the point cloud data to filter out noise points with excessive noise, and only retain valid data Point, improve the accuracy of geometric feature extraction; another example, the vehicle can down-sample the point cloud data, that is, reduce the number of sampling points in the point cloud data to reduce the amount of data, reduce the amount of data processing, and improve the vehicle's positioning speed.
  • noise reduction processing on the point cloud data to be processed, for example, filter out outliers in the point cloud data to filter out noise points with excessive noise, and only retain valid data Point, improve the accuracy of geometric feature extraction
  • the vehicle can down-sample the point cloud data, that is, reduce the number of sampling points in the point cloud data to reduce the amount of data, reduce the amount of data processing, and improve the vehicle's positioning speed.
  • S404 may include the following steps:
  • S4042 Identify at least one object in the point cloud data to be processed.
  • the identification of multiple objects in the point cloud data to be processed in the embodiments of the present application includes, but is not limited to, identifying the attributes of the objects in the point cloud data to be processed, and the attributes may be the categories described for the objects.
  • the category (first category) to which the object belongs includes straight lines, curves, planes, curved surfaces, and so on.
  • the category (the second category) to which the object belongs includes roadsides, roads, tree trunks, the plane of the building, the left edge, the right edge, the upper edge, the lower edge, and the Cross bars, vertical bars, solid lines, dashed lines, turning lines, etc. of road indication lines.
  • the object may not include physical meaning, the vehicle only recognizes its geometric shape, and then uses the technical method corresponding to the geometric shape to extract its geometric features.
  • the object may include specific physical meanings, such as roadsides, road surfaces, road signs, traffic signs, road signs, tree trunks, building planes, left edges, and right edges. , The upper edge, the lower edge, the cross bar and the vertical bar of the traffic pole, the solid line, the dashed line, the turning line of the road indicator line, etc., are not limited in the embodiment of the present application.
  • S4044 Determine the geometric feature of the first object according to the point cloud data of the first object, where the first object is any one of the multiple identified objects.
  • the geometric shape of the object in the point cloud data to be processed is recognized. Further, the point cloud data of each object can be divided from the point cloud data to be processed, and further, the point cloud data of each object The cloud data is fitted with straight lines/curves/planes/surfaces, such as straight lines/curves/planes/surfaces.
  • the geometric characteristics of the geometric expression of the object are the geometric characteristics of the object.
  • the point cloud data of the first object is fitted with a straight line to obtain the first geometric feature of the first object, and the first geometric feature of the first object is the geometric feature of the fitted line ;
  • curve fitting is performed on the point cloud data of the first object to obtain the first geometric feature of the first object, and the first geometric feature of the first object is the geometry of the curve obtained by fitting Features;
  • the geometric shape of the first object is a plane, perform plane fitting on the point cloud data of the first object to obtain the first geometric feature of the first object, and the first geometric feature of the first object is the fitted plane Geometric features; if the geometric shape of the first object is a curved surface, perform surface fitting on the point cloud data of the first object to obtain the first geometric feature of the first object, and the first geometric feature of the first object is obtained by surface fitting The geometric characteristics of the curve.
  • the vehicle can determine the fitting method adopted by each category according to the corresponding relationship between the category and the geometric shape.
  • the linear fitting method is used to calculate them.
  • Geometric Features For another example, for planar objects, such as the plane of a building, the surface of a billboard, etc., a plane fitting method is used to calculate its geometric characteristics.
  • FIG. 4B Please refer to Figure 4B and Figure 4C for the schematic diagrams of extracting the geometric features of the road surface and the road edge.
  • the distribution of the point cloud data of the road edge is shown in Figure 4B.
  • the road surface is usually a flat plane, and the road edge can include the upper edge and the edge.
  • the lower edge where the point cloud data at the upper edge and the point cloud data at the lower edge both have a sudden change in height.
  • the laser beam emitted by the lidar such as the laser beam i in Figure 4B or the laser beam j in Figure 4C, crosses the road and the curb. Therefore, each laser beam
  • the point cloud data obtained by scanning can be observed to change the height.
  • the vehicle can determine the boundary (lower edge) of the road surface and the curb and the upper edge of the curb according to the change in height of the point cloud data obtained by each laser beam. For example, in the case where the height difference ⁇ h between the upper edge and the lower edge is greater than the first height, it is determined that the area between the upper edge and the lower edge is the road edge, and further, based on the point cloud data of the upper edge and the lower edge , Get the geometric characteristics of the curb.
  • the geometric feature of the road surface may be the normal vector of the plane where the road surface is located, as shown in V pavement in FIG. 4B and FIG. 4C.
  • the geometric features of the straight line or curve where the upper/lower edges of the roadside are located can be used as the geometrical characteristics of the roadside.
  • the vector V curb_up or the vector V curb_low is The geometric features of the curb.
  • the combination of multiple coordinate points or the combination of normal vectors on multiple coordinate points on the curve where the boundary between the road surface and the road edge is located (that is, the curve f curb_low in the figure) is used as the geometric feature of the curve, namely (a point Coordinates, normal vector v a of point a, coordinate of point b , normal vector of point b v b, coordinate of point c, normal vector of point c v c ).
  • the average value of the geometric characteristics of the straight line/curve where the upper edge of the road edge is located and the geometric characteristics of the straight line/curve where the lower edge is located can also be used as the geometric feature of the road edge, namely ,
  • the geometric feature of the road edge is (V curb_low +V curb_low )/2, which is not limited in the embodiment of the present application.
  • the geometric feature of the curb may also include the position of the curb, where the position of the curb may be the coordinates of any position in the point cloud data of the curb as the position of the geometric feature corresponding to the object . It should be understood that other rules may also be included to determine the position of the roadside, which is not limited here.
  • the vehicle may also mark the attributes of the geometric feature, such as the size, and the classification.
  • the vehicle can extract the points in the point cloud data of the road where the intensity of the reflection of the laser will be greater than the preset intensity to obtain the point cloud data of the road indicator line, and further, obtain the geometry of the road indicator line according to the point cloud data of the road indicator line feature.
  • FIG. 4D a schematic diagram of the principle of extracting the geometric characteristics of the geometric characteristics of the road indicating line.
  • the geometric characteristics of the road indicating line may be the geometric characteristics of the straight line or the curve where the road indicating line is located.
  • the geometric characteristics of each road indicator line can be as V1-V10 shown in 4D.
  • the road marking line in the point cloud data can be identified based on the image recognition technology, and further, according to the point cloud of the road indication line
  • the data obtains the geometric characteristics of the road indicator line.
  • the vehicle can use image enhancement technology to restore the road indicator line.
  • the wire harness sent by the lidar is blocked by other vehicles due to traffic congestion, resulting in the lack of information about the lane indication line that can be measured.
  • the vehicle can use the geometric characteristics of the lane indication line obtained at the previous moment. And the trajectory of the vehicle to estimate the geometric characteristics of the lane indicator line at the current moment.
  • road indication lines such as turning left, turning right, going straight, and turning around can be used as attributes of geometric features to assist in matching of geometric features.
  • Rod-shaped objects such as traffic poles, tree trunks, road signs or billboards
  • the vehicle can use a three-dimensional grid as a ruler to divide the point cloud data.
  • the size of the three-dimensional grid (X-axis length ⁇ Y-axis length ⁇ Z-axis length) may be 0.2m ⁇ 0.2m ⁇ 0.2m, or other values, there is no limit to this. It should be understood that when point cloud data exists in the three-dimensional grid, there are objects in the three-dimensional grid.
  • the height of the object in the three-dimensional grid is coded to obtain a height distribution map.
  • the three-dimensional grid marked (i, j, k) is represented as the i-th in the X direction, the j-th in the Y direction, and the k-th determined three-dimensional grid in the Z direction.
  • the three-dimensional grid (i , J, K) The encoded height value H is: Wherein, when identifying the existence of an object in the three-dimensional grid (i, j, k), the value of N(k) is 1, otherwise, the value is 0, where i, j, and k are positive integers.
  • Fig. 4E a schematic diagram of the distribution of point cloud data of trees.
  • the encoded height value of the three-dimensional grid (4, 3, 5) is greater than the first threshold.
  • K is 5, and the three-dimensional grid (4, 3, 5)
  • the height value corresponding to the three-dimensional grid (4, 3, 5) is the height of the tree trunk.
  • a straight line fitting is performed, the direction vector of the tree trunk is calculated, and the position of the tree trunk is determined.
  • the point cloud description object is a billboard.
  • the direction vector of the pole of the billboard can be extracted according to the distribution of the point cloud data of multiple heights z located in the three-dimensional grid of (4, 3); for the pole of the billboard, According to the distribution map of the point cloud data in the interval of height Z of 2.2-3.2, the point cloud inside the billboard is extracted, the point cloud inside the billboard is plane-fitted, and the normal vector of the billboard surface is extracted.
  • FIG. 4G a schematic illustration of the geometric characteristics of street light poles, according to the distribution of point cloud data at different heights, it can be recognized that the point cloud description object is a street light pole. Further, according to the three-dimensional (4, 3) Point cloud data in the grid, calculate the direction vector of the street light pole, determine the position of the street light pole, etc.
  • the measured distance is expressed in polar coordinates, as shown in Fig. 4H for a schematic diagram of the distance and angle of the point on the sign Illustrating.
  • the two angles are considered to be the boundary of the object, and the points obtained by combining the adjacent laser beams can segment the measurement points of the object. Further, based on the measurement points obtained by the segmentation, a straight line, a curve, a plane or a curved surface is fitted to obtain the geometric characteristics of each part on the object.
  • Figure 4I shows the distribution diagram of the point cloud data of the building.
  • the vehicle can identify the outer edge of the building (left edge, right edge, upper edge, lower edge, etc.) ) And segment the point cloud data of each outer edge of the building, and then determine the geometric feature of each outer edge as the geometric feature of the straight line/curve where the outer edge is located through straight line/curve fitting and other methods.
  • the vehicle can identify the wall of the building, and then determine the geometric characteristics of the wall of the building by means of plane/curved surface fitting, etc. as the wall of the building. The geometric characteristics of the plane/surface.
  • This embodiment of the application introduces a map generation method, which applies the geometric feature extraction method described in the above embodiment (1), which can be executed by the geometric feature map generation device 140 in FIG. 2, please refer to the geometric feature shown in FIG.
  • a method for generating a map which may include but is not limited to the following steps:
  • the second point cloud data may be the point cloud data of the area to be processed in the point cloud map, or it may be one or more frames of points to be processed in the point cloud map.
  • the cloud data may also be the point cloud data obtained after the above-mentioned multi-frame point cloud data is superimposed on the point cloud between frames, which is not limited in the embodiment of the present application. It should be understood that the geometric feature map generating device can divide the point cloud data in the point cloud map according to the area where the point cloud data is located.
  • the geometric features extracted from the second point cloud data are used to generate a geometric feature map.
  • a geometric feature map refers to the related description of the implementation manner of extracting geometric features from the first point cloud data in the foregoing embodiment (1), which will not be repeated here.
  • the geometric feature may also include attributes, such as category, size, and other information.
  • the geometric features in the geometric feature map can be encoded into various storage formats, such as images, XML, text or tables, etc.
  • the geometric feature map can be stored in a geometric feature map generating device or database.
  • a geometric feature map can be stored in an image format. Each pixel in the image has a certain location. When a geometric feature exists at that location, the value of the pixel can be encoded as the vector, size, or category of the existing geometric feature. Wait.
  • the geometric feature map may also include the location and the address corresponding to the location, the road name corresponding to the geometric feature of the road, and so on.
  • the geometric feature map can be used as a layer in a map (such as a Google map, a Baidu map, a Gaode map, or a point cloud map), or can be used as a separate map, which is not limited.
  • the geometric feature map obtained by this method replaces a large amount of point cloud data on the object through vectors, positions and attributes. Compared with the original point cloud map, the geometric feature map occupies a smaller storage space and greatly reduces the data.
  • the storage capacity of the vehicle can reduce the calculation complexity of subsequent vehicle positioning process data and meet the real-time requirements of vehicle positioning.
  • the first geometric feature extracted from the first point cloud data collected by the vehicle (shown by the dashed line)
  • the geometric feature) and the second geometric feature matching the first geometric feature in the geometric feature map is essentially the expression of the same object in different coordinate systems.
  • the inaccuracy of the first pose estimated by the vehicle on the current pose results in the difference between the first coordinate system determined according to the first pose and the second coordinate system (ie the world coordinate system) in the geometric feature map, that is, two The two coordinate systems need to be rotated and translated to coincide.
  • the first posture of the vehicle can be adjusted by the difference between the first geometric feature and the second geometric feature collected by the vehicle, and the precise position of the vehicle, that is, the second posture, can be obtained.
  • the positioning method can be executed by a positioning device, where the positioning device can be a vehicle or a computer system in the vehicle; it can also be a terminal that communicates with the vehicle, such as a mobile phone or a tablet computer. It can also be a positioning chip or a positioning device, or a server or cloud.
  • the embodiment of the present application is introduced by taking the execution device as the positioning device as an example.
  • the positioning methods shown in the following embodiment (4), embodiment (5) and embodiment (6) can be implemented based on the positioning method shown in embodiment (3).
  • the method may include but is not limited to the following steps:
  • S72 Acquire the first point cloud data collected by the vehicle through the point cloud collection device.
  • S76 Adjust the first posture of the vehicle according to the N first geometric features to obtain the second posture of the vehicle, where the accuracy of the second posture is higher than that of the first posture.
  • the positioning device may also receive the above N first geometric features sent by the vehicle or the terminal, and the N first geometric features are the first points collected from the vehicle through the point cloud collection device Extracted from cloud data.
  • the above method provides a positioning method.
  • the first pose with low accuracy is corrected by N first geometric features extracted from the first point cloud data collected by the point cloud acquisition device, and the accuracy can be obtained.
  • the second high pose uses geometric features with a small amount of data for positioning, which greatly reduces the amount of data calculations, and makes vehicle positioning less time-consuming and positioning The real-time performance is good.
  • the purpose of the embodiment of the present application is to obtain the position and posture of the vehicle with high accuracy, that is, the second pose in the embodiment of the present application.
  • the accuracy of the first pose is lower than the accuracy of the second pose
  • the first pose is the position and pose of the vehicle with low accuracy, that is, it includes the first position and the first pose.
  • the method may further include, S71, acquiring the first pose of the vehicle, where the method for acquiring the first pose of the vehicle may include, but is not limited to, the following three implementation manners:
  • the positioning device determines the first position of the vehicle through the positioning system, which can be the global positioning system (GPS), Beidou navigation satellite system (BDS), base station positioning (also known as mobile location service) (location based service, LBS)), or indoor positioning system, etc.
  • GPS global positioning system
  • BDS Beidou navigation satellite system
  • BBS base station positioning
  • LBS location based service
  • indoor positioning system etc.
  • the first attitude may be the pitch angle, yaw angle, and roll angle of the vehicle measured by the vehicle through the inertial sensor.
  • the vehicle may determine the first coordinate system based on the first attitude, and then describe through the point cloud in the first coordinate system
  • the first point cloud data collected by the collection device may include accelerometers, angular velocity sensors (such as gyroscopes), magnetic sensors, and so on.
  • the positioning device can use the positioning method in the embodiment of the present application to input the second pose and the control input of the vehicle at the previous moment (k-1) into the dynamic equation of the vehicle to predict the current The first pose at time (k).
  • the positioning device can determine the first pose of the vehicle at time T2 according to the second pose of the vehicle at time T1 and the motion trajectory detected by the inertial sensor on the vehicle from time T1 to time T2, where T2 is the point The moment when the cloud collecting device collects the first point cloud data.
  • the vehicle can collect the first point cloud data of the current environment through the point cloud collection device; and then, send the collected first point cloud data to the positioning device, so that the positioning device
  • execute S74 to obtain N first geometric features.
  • the first point cloud data is the information of the points on the surface of the object observed by the vehicle represented in the space determined by the first pose, and each of the N first geometric features is used to indicate The geometric feature of an object observed by the vehicle in the space determined by the first pose (also referred to as the space described by the first coordinate system).
  • step S74 may also be executed by the vehicle, and further, the vehicle sends the N first geometric features extracted from the first point cloud data to the positioning device.
  • the first coordinate system is a coordinate system (space) determined by the vehicle based on the first pose, and since the accuracy of the first pose is not high, there is a deviation between the first coordinate system and the world coordinate system.
  • N first geometric features are extracted from the first point cloud data as an example.
  • the method may further include:
  • S75 Search for N second geometric features matching the N first geometric features in the geometric feature map, where the first geometric features and the second geometric features are in one-to-one correspondence.
  • the geometric feature map may include the geometric features of the object described in the world coordinate system.
  • the positioning device may adjust the first pose based on the N first geometric features observed by the vehicle and the N second geometric features in the geometric feature map to obtain the second pose.
  • the geometric feature map includes the geometric features of each area on the map.
  • the positioning device caches the geometric feature map.
  • the positioning device can search for N second geometric features matching the N first geometric features from the first area of the geometric feature map, such as In the first geometric feature and the second geometric feature shown in FIG. 6, the first area may be an area determined based on the first pose, and the first area may be larger than the scanning range of the point cloud acquisition device of the vehicle.
  • the positioning device may request the geometric feature map generating device or the geometric feature map generating device to obtain the geometric feature in the first area in the geometric feature map, which is not limited.
  • the positioning device may determine the scene where the vehicle is located at the first position according to the first position and the road at the first position, for example, road intersections, multi-layer roads, etc.
  • the first area is delineated in different ways. Its main realization is as follows:
  • the first area may be an elliptical area 801 as shown in FIG. 8A, and the first location 802 is located in an ellipse. ⁇ 801. It can be understood that the first area is still a spherical area or a circular area determined by the first position as the center and the first length as the radius; it may also be a rectangular parallelepiped area or a rectangular area determined by the first position as the center; it may also be a point.
  • the cloud collection device is in the first position, the detection range of the point cloud collection device.
  • the first area may be as shown in FIG. 8B In the area 803, the first position 804 is located in the area 803. At this time, due to the inaccuracy of the first pose, the first area needs to consider the objects on each road to prevent the first area in the divided geometric feature map from not completely covering the objects described by the N first geometric features .
  • the first area may be the area 805 as shown in FIG. 8C.
  • the positioning device may first determine the road layer where the vehicle is located, and then determine the first area 805. Among them, the positioning device can determine the road where the vehicle is located according to the destination to be reached by the vehicle navigation, or the pose and movement path obtained by the vehicle's previous positioning, the height in the first pose, and so on. Furthermore, the first area can be determined by the method shown in 8A or 8B above.
  • the geometric features in the first region in the geometric feature map can be matched with the N first geometric features extracted from the collected first point cloud data, so as to obtain the same value as the N first point cloud data.
  • the first geometric feature and the second geometric feature are matched one by one.
  • the embodiment of the present application takes the first area including M geometric features, and the N first geometric features extracted by the positioning device from the collected first point cloud data as an example, where M and N are positive integers, and M ⁇ N, the matching process can include but not limited to the following three implementation methods:
  • the positioning device can match the first geometric feature with the M geometric features one by one.
  • the geometric feature with the smallest deviation from the first geometric feature is the second geometric feature matching the first geometric feature (this In the application, it is also referred to as the second geometric feature corresponding to the first geometric feature).
  • the deviation of the first geometric feature and the second geometric characteristic V i U i may be between the second geometric feature vectors a first vector V i of the geometric features of U i ⁇ The included angle.
  • the deviation of the first geometric feature V i and the second geometric feature U i may be the difference between the coordinate point in the first geometric feature V i and the coordinate point in the second geometric feature U i The distance between.
  • the positioning device can achieve matching through the attributes of geometric features.
  • the attributes of the two geometric features match, including that the size error of the two geometric features is less than the first threshold, the objects of the two geometric features belong to the same category, etc., for example, the error of the length of the two geometric features is less than 0.01m; and
  • the categories of objects that two geometric features belong to are traffic poles and so on.
  • the positioning device can achieve matching through the combination of geometric feature attributes and geometric features.
  • the positioning device can also select the geometric features geometric features in the first region than the second threshold value with the deviation of the first geometric feature V i of geometric features if they are selected for a plurality of time can be further Comparing whether the attributes of the two geometric features match, to determine a second geometric feature U i that matches the first geometric feature V i from the selected geometric features.
  • the positioning device can also select the geometric features geometric features in the first region than the second threshold value with the deviation of the first geometric feature V i of geometric features if they are selected for a plurality of time can be further Comparing whether the attributes of the two geometric features match, to determine a second geometric feature U i that matches the first geometric feature V i from the selected geometric features.
  • the positioning device can select different matching implementation methods according to different categories of the first geometric feature.
  • the vectors of the first geometric features can be calculated separately from the first geometric features.
  • the included angle of the vector of each geometric feature in the geometric features in the region, and further, among the geometric features in the first region, the geometric feature with the smallest included angle with the first geometric feature is the second matching the first geometric feature.
  • the positioning device obtains the first position in real time, loads the geometric features in the first region in real time, and matches the geometric features in the first region with the N first geometric features.
  • N first geometric features are taken as an example to illustrate.
  • the N first geometric features can be selected by the positioning device from all the geometric features extracted from the first point cloud data collected by the point cloud collection device.
  • N first geometric features may also be all geometric features extracted from the first point cloud data collected by the point cloud collection device.
  • the specific implementation of S76 may include, but is not limited to, the five positioning modes described in the following embodiment (4), embodiment (5), and embodiment (6).
  • an implementation of S76 may include the following steps:
  • the positioning device determines the transformation relationship between the geometric features according to the N first geometric features and the N second geometric features in the geometric feature map.
  • S762 Adjust the first pose of the vehicle according to the transformation relationship between the geometric features to obtain the second pose.
  • the first pose with low accuracy is corrected by the transformation relationship between the observed first geometric feature and the second geometric feature in the geometric feature map, and a second pose with high accuracy can be obtained.
  • the embodiment of the present application adopts geometric features with a small amount of data for registration and positioning, which greatly reduces the amount of data calculations, so that vehicle positioning is less time-consuming and has good real-time positioning.
  • the first positioning method :
  • the positioning principle of the first positioning method is: through the transformation relationship between the first coordinate system determined based on the current first pose of the vehicle and the second coordinate system adopted by the geometric feature map (that is, also referred to in the embodiments of the present application)
  • the target transformation amount transforms the first pose of the vehicle (that is, the estimated pose) to obtain the second pose of the vehicle, which is the accurate pose of the vehicle.
  • the target conversion amount between the first coordinate system and the second coordinate system that is, the conversion of the first coordinate system to the second coordinate system requires rotation and translation.
  • the rotation and translation can be calculated by calculating the first geometric feature and the first geometric feature.
  • the rotation and translation between the second geometric features corresponding to the features are obtained.
  • the positioning device can extract the N first geometric features from the collected point cloud data to obtain N third geometric features after rotating R and translating t, so that the N
  • the conversion amount composed of the rotation R and the translation t with the smallest error between the third geometric feature and the N second geometric features is the inverse of the target conversion amount, and N is a positive integer greater than 1.
  • the specific implementation of the first positioning manner may include, but is not limited to, some or all of the following steps:
  • S912 Transform the N first geometric features by the first transformation amount to obtain N third geometric features, and the third geometric features have a one-to-one correspondence with the first geometric features.
  • S914 Sum the N errors to obtain a first error, where the first error is obtained by direct summation or weighted summation of the N errors.
  • S915 Determine whether the number of iterations or the first error satisfies the iteration stop condition.
  • the positioning device can determine whether the number of iterations is equal to a preset number.
  • the preset number can be 4, 5, 10, 30, or other values. If so, execute S916; otherwise, , Go to S917.
  • the positioning device may determine whether the first error has converged, and if so, execute S916; otherwise, execute S917.
  • the embodiment of the present application may also include other implementation manners, for example, judging whether the first error is less than a preset value, such as 0.1, 0.2, or other values, which is not limited.
  • the positioning device After S916, the positioning device repeatedly executes S912-S915 until the number of iterations or the first error satisfies the stop iteration condition.
  • S917 The output first transformation amount T, the first transformation amount is the first target transformation amount.
  • the error ie, the first error
  • the process of adjusting the initial transformation amount is the process of reducing this first error as much as possible.
  • the first transformation amount is the initialized rotation and translation preset by the positioning system. Through multiple iterations and adjustments, the first error becomes smaller and smaller.
  • the transformation amount that minimizes the first error is the first target transformation amount. It should also be understood that, in another implementation of the embodiment of the present application, the first target transformation amount may also be a transformation amount obtained by adjusting a preset number of times.
  • a first objective function may be constructed, and the first objective function may include but is not limited to the following three forms:
  • the positioning device may determine the first error according to the first objective function, where the first objective function may be:
  • a first error ⁇ conversion amount comprises a first rotation R and translation t
  • w i is the weight of the first geometric feature weights V i
  • U i for the second geometric feature of the first geometric feature corresponding to V i
  • the index of the first geometric feature in a geometric feature, i is a positive integer, i ⁇ N.
  • the first objective function can be expressed as:
  • the rotation R and the translation t are the first transformation amount, which is a variable; i is the index of the first geometric feature among the N first geometric features, i is a positive integer, and i ⁇ N.
  • the value of the first objective function is continuously calculated by adjusting the rotation R and the translation t, so that the rotation R and the translation t with the smallest first objective function are the first target transformation amount.
  • the rotation R and the translation t that minimize the first objective function can also be obtained by factorization and other methods, which are not limited in the embodiment of the present application.
  • (Rv i + t) is the second geometric feature vector u i U i of the third characteristic geometric transform by the initial conversion amount.
  • w i is the weight of the first geometric feature weights V i, a first geometric feature for limiting the contribution of V i conversion amount of the first target.
  • the heavy weight of the first geometric features with respect to the device according to the first geometric feature from the point cloud on a vehicle or to determine the acquisition for example, the right to a first geometric feature weights w i V i and V i belongs to the first geometric feature
  • the distance of the object relative to the vehicle is negatively correlated, that is, the first geometric feature that is closer to the point cloud collection device or the vehicle has a greater weight.
  • the weight of the first geometric feature can also be determined according to the type of object corresponding to the first geometric feature, that is, different weights can be set for different objects, for example, For a building plan, the normal vector of the building plan obtained by it has high accuracy, and its corresponding first geometric feature can have a higher weight; for another example, for a tree trunk, the accuracy of the direction vector of its corresponding straight line If it is lower, its corresponding first geometric feature may have a lower weight. It should be understood that it is not limited to the above-mentioned weight setting method.
  • the embodiment of the present application may also use other setting methods to set the weight, for example, comprehensively considering the distance of the first geometric feature and the type of object corresponding to the first geometric feature;
  • the application embodiment may also not include the weight, that is, for any one of the first geometric features, the weight is 1, which is not limited here.
  • the first geometric feature may include a vector and a position.
  • the first objective function may be expressed as:
  • G 1, i is the position of the first geometric feature of V i
  • G 2, i is the position U i of the second geometric feature
  • W i is a weight vector v i first geometric feature V i in weight, particular reference may be The relevant description of the first form of the first objective function will not be repeated here
  • V i position of the first geometric feature of G 1, i is the right weight, for, defining a first geometric feature contributions i i V G 1 position conversion amount of the first target.
  • the weight of the position of the first geometric feature can be determined by the type of the corresponding object, that is, different weights can be set for different objects, for example, for a traffic pole, its position can be accurately determined according to the position setting rules ,
  • the position of the geometric feature has a higher weight, such as 1.
  • the position of the geometric feature has a lower weight, such as 0.
  • the weight can also be set through other setting methods.
  • the embodiment of the present application does not limit it.
  • the second positioning method is a first positioning method
  • the transformation relationship is explained by taking the transformation of the first geometric feature into the second geometric feature in the geometric feature map as an example. It should be understood that the transformation relationship can also be the transformation of the second geometric feature in the geometric feature map. The geometric feature is transformed into the first geometric feature.
  • the difference is the calculation method of the transformation relationship.
  • the positioning device can obtain N fourth geometric features by rotating R and translating t respectively through the N second geometric features, so that the N fourth geometric features are extracted from the collected point cloud data
  • the obtained rotation R′ and translation t′ with the smallest error between the N first geometric features are the inverse matrix of the target conversion amount T, and N is a positive integer greater than 1.
  • the specific implementation of the second positioning manner may include, but is not limited to, some or all of the following steps:
  • S922 Transform the multiple second geometric features by the second transformation amount to obtain multiple fourth geometric features, and the fourth geometric features have a one-to-one correspondence with the second geometric features.
  • S924 Sum the N errors to obtain a second error, where the second error is obtained by direct summation or weighted summation of the N errors.
  • S925 Determine whether the number of iterations or the second error satisfies the iteration stop condition. For specific implementation, refer to the relevant description in S914 above. If the judgment result is yes, then execute S926; otherwise, execute S927.
  • the positioning device After S926, the positioning device repeatedly executes S922-S925 until the number of iterations or errors meet the conditions for stopping iterations.
  • a second objective function can be constructed to realize the calculation of the second error. Similar to the first form of the first objective function, the second objective function can be expressed as:
  • the second objective function can be expressed as:
  • w i , u i, v i, G 1 , i, G 2, i, i may refer to the related description of the first embodiment in the positioning related description is not repeated here.
  • first objective function or the second objective function may also include other forms, and the error may also be a mean absolute error (MAE). , Mean Squared Error (MSE), Root Mean Squared Error (RMSE) or other forms, etc., are not limited in this application.
  • MSE Mean Squared Error
  • RMSE Root Mean Squared Error
  • the calculation process of the target conversion amount is the solution of the minimization of the first objective function or the second objective function
  • the method of minimizing the objective function may include, but is not limited to, Gauss-Newton method, gradient descent method, LM (Levenberg- Marquardt) method, QR decomposition or other solving methods.
  • the method may also include S71 and S75.
  • S76 may include but Not limited to the following steps:
  • S763 Determine the predicted pose of the vehicle at the current moment according to the second pose at the previous moment and the control parameters of the vehicle at the previous moment, where the previous moment is a moment before the current moment.
  • S764 Update the predicted pose of the vehicle according to the error between the observed value of the second parameter and the predicted value of the second parameter to obtain the second pose of the vehicle.
  • the observed value of the second parameter is determined based on the first geometric feature observed by the vehicle in the first pose
  • the predicted value of the second parameter is determined based on the predicted pose and the second geometric feature in the geometric feature map .
  • the positioning device determines the first position of the vehicle according to the positioning system and according to the inertial sensor.
  • the first posture of the vehicle is determined, and the first posture includes a first position and a first posture.
  • the positioning device can also obtain N first geometric features observed by the vehicle, and one first geometric feature is used to represent in the space determined by the first pose
  • the geometric feature of an object observed by the vehicle N is a positive integer
  • the error between the observed value of the second parameter and the predicted value of the second parameter is determined based on the first pose, the predicted pose, N first geometric features, and N second geometric features.
  • the positioning device searches the geometric feature map for N second geometric features that match the N first geometric features.
  • the geometric features please refer to the relevant descriptions of step S75 and the three implementations of the matching process in the above-mentioned embodiment (3), which will not be repeated here.
  • Embodiment (5) provides a vehicle positioning method.
  • the estimated pose is scored by the observed first geometric feature and the second geometric feature in the geometric feature map, and the estimated pose with the highest score is determined as the actual position of the vehicle.
  • the embodiment of the present application uses geometric features with a small amount of data for positioning, which greatly reduces the amount of data calculation, so that vehicle positioning is less time-consuming and has good real-time positioning.
  • Step S76 may specifically calculate the second pose of the vehicle through the Kalman filtering method.
  • FIG. 10B a schematic flow diagram of the third positioning method and a schematic explanatory diagram of the third positioning method shown in FIG. 10C.
  • the third positioning method includes the Kalman prediction process and the Kalman update process.
  • the third positioning method is implemented Including but not limited to the following steps:
  • S102 is the Kalman prediction process, which may be an implementation manner of the above S71, in which case the first pose is the predicted pose.
  • the state of the vehicle can be obtained according to the dynamic equation of the vehicle (ie, the equation of motion of the clock in the embodiment of the present application).
  • the state of the vehicle is a pose, including a position and a posture.
  • the embodiment of the present application uses two-dimensional coordinates To indicate the position of the vehicle, with the yaw angle To indicate the yaw angle of the vehicle.
  • the position of the vehicle can also be represented by three-dimensional coordinates, and the attitude of the vehicle can be represented by three angles (pitch angle, yaw angle, and roll angle). Not limited.
  • the pose prediction is carried out through the above-mentioned vehicle motion equation.
  • the motion equation of the vehicle (that is, the prediction equation in the embodiment of this application) can be expressed as:
  • the covariance matrix is Q k .
  • the predicted pose can be And N second geometric features are input into the observation equation to obtain the predicted value of the second parameter at the current moment (k).
  • the observation equation is based on the equation of the second parameter, and the second parameter may be at least one of the distance, azimuth angle, and altitude angle of the object described by the first geometric feature relative to the own vehicle.
  • the predicted value of the second parameter is the second parameter of the N second geometric features respectively relative to the vehicle in the predicted pose.
  • the observed value of the second parameter is the second parameter of the N first geometric features respectively relative to the vehicle in the first pose.
  • the distance and azimuth angle of the object with respect to the vehicle described by the second parameter as the first geometric feature are taken as an example for illustration. It should be understood that Kalman filtering can also be used to construct other second parameters and observation equations. It should be understood that the object described by the first geometric feature and the object described by the second geometric feature corresponding to the first geometric feature are the same object.
  • the second parameter may be the distance and azimuth angle of the location point relative to the vehicle itself.
  • the predicted value of the second parameter at the current moment (k) can be expressed as: Wherein the first geometric feature parameter V i with respect to a second vehicle at the predicted pose, i.e., the predicted value of the second parameter of the first geometric feature corresponding to V i for:
  • G 2,i [x 2,i ,y 2,i ] represents the coordinate point in the second geometric feature U i .
  • the embodiment of this application does not limited.
  • the second parameter may be the distance, azimuth, and height angle of the vector relative to the own vehicle. If the N first geometric features include the direction of the object, that is, expressed by the direction vector, the predicted value of the second parameter at the current moment (k) can be expressed as, At this time, a first characteristic geometrical parameter V i with respect to a second pose at the prediction of the vehicle, i.e., the predicted value of the second parameter of the first geometric feature corresponding to V i for:
  • Is the predicted distance at the current moment Predict the azimuth for the current moment, Is the predicted altitude angle at the current moment.
  • the distance OA1 from the origin of the vehicle coordinate system to the vector is The projection of coordinate A1 to the OXY plane of the vehicle coordinate system is coordinate A2, then the angle between OA2 and OX is The angle between OA2 and OA1 is
  • the positioning device may input the first pose S1 and the N first geometric features into the observation equation to obtain the observation value of the second parameter at the current moment (k).
  • the first pose can be expressed as (x int , y int ) represents the first position of the vehicle, and ⁇ int represents the first posture of the vehicle.
  • the predicted value of the second parameter of the N first geometric features at the current moment can be expressed as
  • the third error at the current moment can be expressed for
  • S110 is the Kalman update process.
  • the update minimizes the state error, and the equation for the pose update can be obtained as:
  • the update equation of Kalman gain is:
  • the update equation of the covariance matrix of the pose is: Among them, R k and Q k are the observation noise matrix and the pose noise matrix respectively.
  • a k and H k are respectively the vehicle pose conversion matrix and the observation conversion matrix. After update That is the second pose at the current moment.
  • the positioning device can calculate the second pose at the current moment Input to the motion equation and predict the predicted pose at the next moment (k+1) The second pose, and further, by the method described in S71-S76 or S102-S110 above, the second pose of the vehicle at the next moment (k+1) is obtained
  • the method may also include S71 and S75, where S76 may include but is not limited to the following steps:
  • S705 Estimate the pose of the vehicle according to the first pose, and obtain multiple sets of estimated poses.
  • S706 Determine the scores of multiple sets of estimated poses according to the N first geometric features and the N second geometric features in the geometric feature map.
  • S707 Determine the second pose of the vehicle according to the score of each of the multiple sets of estimated poses, where the score of the first set of estimated poses is used to indicate the first set of estimated poses and the second set of poses.
  • the degree of proximity of the pose, the first set of estimated poses is any one of the estimated poses among multiple sets of estimated poses.
  • the score of the first set of estimated poses is determined based on the first set of estimated poses, the first pose, the N first geometric features observed by the vehicle, and the N second geometric features in the geometric feature map. .
  • the positioning device may also obtain N first geometric features observed by the vehicle, and one first geometric feature is used to represent in the space determined by the first pose
  • the geometric feature of an object observed by the vehicle N is a positive integer; further, search for N second geometric features matching the N first geometric features in the geometric feature map, the first geometric feature and the second geometric feature are one One correspondence; further, the score of each set of estimated poses is determined based on each set of estimated poses, first poses, N first geometric features, and N second geometric features.
  • the positioning device searches the geometric feature map for those matching the N first geometric features.
  • the N second geometric features please refer to the relevant descriptions of step S75 and the three implementation manners of the matching process in the foregoing embodiment (3), which will not be repeated here.
  • the error between the observed value of the second parameter determined by the observed first geometric feature and the predicted value of the second parameter determined based on the second geometric feature in the geometric feature map is relative to the current moment
  • the predicted pose is updated to obtain the actual pose of the vehicle.
  • the positioning device adjusts the first pose to obtain the second pose at least through the following three positioning methods:
  • a priori pose estimation can be performed on the pose of the vehicle to obtain multiple sets of estimated poses and the estimated value of the first parameter in each set of estimated poses; further, for each estimated pose For the pose, a first score is obtained according to the error between the estimated value of the first parameter and the observed value of the first parameter, and the degree of closeness to the second pose is evaluated by the first score. It should be understood that the smaller the error between the estimated value and the observed value corresponding to the estimated pose, the higher the first score of the estimated pose, and the closer the estimated pose is to the actual pose of the vehicle, that is, the second pose.
  • FIG. 11B is a schematic flowchart of a fourth positioning method provided by an embodiment of this application. The fourth positioning method may include but is not limited to the following steps:
  • S11011 Perform a priori estimation on the second pose S2 of the vehicle according to the first pose S1, and obtain the estimated poses of group D, where D is a positive integer greater than 1.
  • the first pose S1 may be the predicted pose at the current moment predicted by the positioning device according to the second pose at the previous moment and the vehicle's motion equation, which is described in the third positioning method above of Please refer to the related description in the third positioning method above, which will not be repeated here. It should be understood that the first pose may also be the pose at the current moment obtained through GPS and the vehicle's inertial sensor positioning, or a pose obtained by positioning in other ways, which is not limited in the embodiment of the present application.
  • the estimated pose of group D may be a set of poses that are normally distributed around the first pose S1 as the expectation of the first pose S1.
  • D can be 100, 1000, 3000 or other values.
  • the estimated pose of group D can be expressed as
  • S11012 Determine an estimated value of the first parameter corresponding to each set of estimated poses according to each set of estimated poses and N second geometric features.
  • the posture is estimated
  • the positioning device can use the estimated pose And N second geometric features are input to the observation equation to get the estimated pose The corresponding estimated value of the first parameter Estimated pose One of the estimated poses for the D group.
  • the observation equation is based on the equation of the first parameter.
  • the first parameter may be the distance, azimuth, and height angle of the object described by the first geometric feature relative to the own vehicle. At least one of them.
  • the estimated value of the first parameter is the first parameter of the N second geometric features respectively relative to the vehicle in the estimated pose.
  • the observed value of the first parameter is the first parameter of the N first geometric features with respect to the vehicle in the first pose.
  • step S104 For related descriptions of the kinematics equations and observation equations of the vehicle, please refer to the related descriptions in the above-mentioned embodiment (5), which will not be repeated here.
  • the calculation method of the estimated value of the first parameter corresponding to the estimated pose please refer to the related description in the calculation method of the predicted value of the second parameter in the above embodiment (5) (ie, step S104), which will not be repeated here.
  • the estimated values of the first parameters corresponding to the D groups of estimated poses can be obtained through the above S11012.
  • the estimated value of the first parameter corresponding to the estimated pose of group D can be expressed as:
  • S11013 Determine the observed value of the first parameter according to the first pose and the N first geometric features.
  • S11014 Determine the first score of each group of estimated poses according to the error between the estimated value of the first parameter corresponding to each set of estimated poses and the observed value of the first parameter.
  • the estimated pose score in S1104 is also referred to as the first estimated pose score in the fourth positioning manner.
  • the estimated pose is the second pose (the actual pose). If the estimated value of the first parameter corresponding to the estimated pose is equal to the observed value of the first parameter, the estimated pose is the second pose.
  • the function for calculating the first score may be constructed according to the error between the estimated value of the first parameter corresponding to the estimated pose and the observed value of the first parameter. The higher the first score, the closer the estimated pose corresponding to the first score is to the second pose. For example, estimated pose The first score of can be calculated by the following formula:
  • corr() is the Pearson product-moment correlation coefficient, which can be the estimated pose The corresponding estimated value of the first parameter The quotient of the covariance with the observed value Z k of the second parameter and the product of the two standard deviations.
  • S11015 Obtain a second pose according to the estimated poses of the D group and the first score of each estimated pose in the D estimated poses.
  • the second pose may be the estimated pose with the highest first score among the highest estimated poses in the D group.
  • the second pose S2 can be expressed as:
  • ⁇ j is the normalization coefficient of score j.
  • the second pose may also be obtained in other ways based on the estimated poses in the D group and the score of each estimated pose in the D estimated poses, which is not limited in the embodiment of the present application.
  • FIG. 11C is a schematic flowchart of a fifth positioning method provided by an embodiment of this application.
  • the fifth positioning method may include but is not limited to the following steps:
  • S11021 Perform a priori estimation on the second pose S2 of the vehicle according to the first pose S1 to obtain the estimated poses of group D, where D is a positive integer greater than 1.
  • N second geometric features are respectively transformed through the transformation relationship between each set of estimated poses and the first poses to obtain N fifth geometric features corresponding to each set of estimated poses, and N second The geometric features correspond to the N fifth geometric features one-to-one.
  • the second geometric feature corresponds to the fifth geometric feature one to one.
  • the second geometric feature and the fifth geometric feature are expressions for the same object in different coordinate systems.
  • estimated pose I the second pose of the vehicle (the actual pose).
  • the estimated pose The corresponding N fifth geometric features are the same as the N first geometric features.
  • the degree of proximity to the second pose is the estimated pose The score (also referred to as the second score in the embodiments of this application).
  • the second score is used to evaluate how close the estimated pose is to the second pose. The higher the second score, the closer the estimated pose corresponding to the second score is to the second pose.
  • S11023 Determine the second score of each set of estimated poses according to the errors between the N fifth geometric features and the N first geometric features corresponding to each set of estimated poses.
  • N 1 first geometric features are represented by vectors
  • N 2 first geometric features are represented by coordinate points
  • m is the index of the first geometric features in the N 1 first geometric features
  • n Is the index of the first geometric feature in the N 2 first geometric features
  • m, n, N 1 , and N 2 are positive integers
  • m ⁇ N 1 , n ⁇ N 2 , N 1 +N 2 N
  • ⁇ w m represents the weight of the first geometric feature U m
  • d w n represents the weight of the first geometric feature U n
  • the weight of the first geometric feature can be based on the distance between the first geometric feature and the vehicle
  • ⁇ V m , j O m > represents the angle between the two vectors, and the corresponding weight coefficient is ⁇ w m
  • the feature is the target In the position,
  • estimate the pose The score can be expressed as:
  • S11024 Obtain the second pose according to the second score of the estimated pose of each group in the estimated pose of the D group and the estimated pose of the D group.
  • S11024 can be referred to the related description in S11014 in the above-mentioned fourth positioning mode, which will not be repeated here.
  • FIG. 11D is a schematic flowchart of a sixth positioning method provided by an embodiment of this application.
  • the sixth positioning method may include but is not limited to the following steps:
  • S11031 Perform a priori estimation on the second pose S2 of the vehicle according to the first pose S1 to obtain the estimated poses of group D, where D is a positive integer greater than 1.
  • N first geometric features are respectively transformed through the transformation relationship between each set of estimated poses and the first poses to obtain N sixth geometric features corresponding to each set of estimated poses, N first The geometric features correspond to the N sixth geometric features one-to-one.
  • the first geometric feature and the sixth geometric feature are for the expression of the same object in different coordinate systems.
  • estimated pose I the second pose of the vehicle (the actual pose).
  • the estimated pose The corresponding N fifth geometric features are the same as the N second geometric features.
  • the degree of proximity to the second pose is the estimated pose The score (also called the third score).
  • the third score is used to evaluate how close the estimated pose is to the second pose. The higher the third score is, the closer the estimated pose corresponding to the third score is to the second pose.
  • S11033 Determine the third score of each set of estimated poses according to the errors between the N sixth geometric features and the N second geometric features corresponding to each set of estimated poses.
  • the estimated pose in S11033 The calculation method of the error between the corresponding N sixth geometric features and the N second geometric features can refer to the estimated pose in the fifth positioning method mentioned above.
  • the calculation method of the error between the corresponding N fifth geometric features and the N first geometric features ie, step S11023 is described in related descriptions, which will not be repeated here.
  • S10134 Obtain the second pose according to the third score of each group of estimated poses in the estimated poses of the D group and the estimated poses of the D group.
  • the above positioning method can be applied to scenes that require accurate estimation of the pose of the vehicle, such as vehicle navigation and automatic driving of the vehicle. Through the above positioning method, not only can the vehicle provide a more accurate position, but also the vehicle can be obtained. Stance.
  • FIG. 12 is a schematic block diagram of a positioning device in an embodiment of the present invention.
  • the positioning device shown in FIG. 12 (the device 1200 may specifically be the positioning device in the embodiment corresponding to FIG. 2, such as a vehicle 160/terminal 180/positioning server 190, it may also be the vehicle 100 in FIG. 3.
  • the positioning device 1200 may include:
  • the first obtaining unit 1201 is used for the first point cloud data collected by the vehicle through the point cloud collecting device;
  • the feature extraction unit 1202 is configured to extract N first geometric features from the first point cloud data, where N is a positive integer;
  • the adjustment unit 1203 is configured to adjust the first pose of the vehicle according to the N first geometric characteristics to obtain the second pose of the vehicle, and the accuracy of the second pose is higher than that of the first pose.
  • the device 1200 further includes:
  • the second acquiring unit 1204 is configured to acquire the first position of the vehicle before the adjustment unit adjusts the first position of the vehicle according to the N first geometric characteristics to obtain the second position of the vehicle posture.
  • the device 1200 further includes:
  • the matching unit 1205 is configured to search for N second geometric features matching the N first geometric features in the geometric feature map.
  • the positioning device may not include the first acquisition unit 1201 and the feature extraction unit 1202 in the positioning device 1200 described above, and the adjustment in the positioning device 1200 may include a receiving unit for receiving a vehicle. Or N first geometric features sent by the terminal.
  • the second acquiring unit 1204 and the matching unit 1205 in the positioning device 1200 are not necessary units of the positioning device 1200. It should also be noted that the above positioning device 1200 also includes other units used to implement the positioning method described in the embodiment (3), embodiment (4), embodiment (5) or embodiment (6). For the specific implementation of each unit or other units of the device 1200, refer to the relevant descriptions in the above-mentioned embodiment (3), embodiment (4), embodiment (5) or embodiment (6), and will not be repeated here.
  • FIG. 13 is a schematic diagram of the hardware structure of a positioning device provided by an embodiment of the present application.
  • the positioning apparatus 1300 shown in FIG. 13 includes a memory 1301, a processor 1302, a communication interface 1303, and a bus 1304.
  • the memory 1301, the processor 1302, and the communication interface 1303 implement communication connections between each other through the bus 1304.
  • the memory 1301 may be a read only memory (Read Only Memory, ROM), a static storage device, a dynamic storage device, or a random access memory (Random Access Memory, RAM).
  • the memory 1301 may store a program. When the program stored in the memory 1301 is executed by the processor 1302, the processor 1302 and the communication interface 1303 are used to execute each step of the positioning method of the third embodiment of the present application.
  • the processor 1302 may adopt a general-purpose central processing unit (Central Processing Unit, CPU), a microprocessor, an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a graphics processing unit (graphics processing unit, GPU), or one or more
  • the integrated circuit is used to execute related programs to realize the functions required by the units in the positioning device 1200 of the embodiment of the present application, or to execute the positioning method of the third embodiment of the method of the present application.
  • the processor 1302 may also be an integrated circuit chip with signal processing capability. In the implementation process, each step of the sample generation method of the present application can be completed by an integrated logic circuit of hardware in the processor 1302 or instructions in the form of software.
  • the aforementioned processor 1302 may also be a general-purpose processor, a digital signal processor (Digital Signal Processing, DSP), an application specific integrated circuit (ASIC), a ready-made programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic devices , Discrete gates or transistor logic devices, discrete hardware components.
  • DSP Digital Signal Processing
  • ASIC application specific integrated circuit
  • FPGA ready-made programmable gate array
  • the methods, steps, and logical block diagrams disclosed in the embodiments of the present application can be implemented or executed.
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like.
  • the steps of the method disclosed in the embodiments of the present application may be directly embodied as being executed and completed by a hardware decoding processor, or executed and completed by a combination of hardware and software modules in the decoding processor.
  • the software module can be located in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, or electrically erasable programmable memory, registers.
  • the storage medium is located in the memory 1301, and the processor 1302 reads the information in the memory 1301, and combines its hardware to complete the functions required by the units included in the positioning device 1200 of the embodiment of the present application, or perform the third embodiment of the present application, The positioning method in embodiment (4), embodiment (5) or embodiment (6).
  • the communication interface 1303 uses a transceiving device such as but not limited to a transceiver to implement communication between the device 1300 and other devices or a communication network.
  • a transceiving device such as but not limited to a transceiver to implement communication between the device 1300 and other devices or a communication network.
  • data such as point cloud data, first geometric features, second geometric features, and geometric feature maps can be acquired through the communication interface 1303.
  • the communication interface 1303 is also used to implement communication with other devices, such as a geometric feature map generating device, a map server, and a terminal.
  • the bus 1304 may include a path for transferring information between various components of the device 1300 (for example, the memory 1301, the processor 1302, and the communication interface 1303).
  • the vehicle When the positioning device 1300 is installed on a vehicle, the vehicle may be the vehicle 100 shown in FIG. 3, and the device 1300 may also include a point cloud acquisition device 1305, and the positioning device 1300 may also perform geometric feature extraction in the first embodiment. method.
  • the point cloud acquisition device 1305 may be a laser radar (laser radar), a stereo camera (stereo camera), a time of flight camera (time of flight camera) and other devices that can acquire point cloud data.
  • the positioning device 1300 is a terminal, such as a mobile phone, a tablet computer, or a server, the cloud, etc.
  • the point cloud collection device 1305 is not a necessary component.
  • first acquiring unit 1201, the second acquiring unit 1204, and the receiving unit in the positioning device 1200 may be equivalent to the communication interface 1303 in the device 1300, and the extracting unit 1202, the adjusting unit 1203, and the matching unit 1205 may be equivalent to the processor 1302. .
  • Figure 14 is a schematic block diagram of a geometric feature extraction device in an embodiment of the present invention.
  • the geometric feature extraction device shown in Figure 14 (the device 1400 may specifically be the positioning device in the embodiment corresponding to Figure 2, such as a vehicle 160/ The terminal 180/positioning server 190, or the apparatus 1400 may be a geometric feature map generating device 140).
  • the apparatus 1400 may include:
  • the obtaining unit 1401 is configured to obtain the point cloud data to be processed
  • the extraction unit 1402 is configured to extract at least one geometric feature from the point cloud data to be processed; wherein, the at least one geometric feature is used for vehicle positioning.
  • the device 1400 may further include a map generating unit, configured to generate a geometric feature map according to the extracted geometric features.
  • a map generating unit configured to generate a geometric feature map according to the extracted geometric features.
  • the device 1400 also includes other units used to implement the geometric feature extraction method described in embodiment (1) or the unit in the geometric feature map generation method described in embodiment (2).
  • the device 1400 also includes other units used to implement the geometric feature extraction method described in embodiment (1) or the unit in the geometric feature map generation method described in embodiment (2).
  • Each of the above-mentioned device 1400 For the specific implementation of the unit or other units, please refer to the related description in the above-mentioned embodiment (1) or embodiment (2), which will not be repeated here.
  • FIG. 15 is a schematic diagram of the hardware structure of a geometric feature extraction device provided by an embodiment of the present application.
  • the apparatus 1500 for extracting geometric features shown in FIG. 15 includes a memory 1501, a processor 1502, a communication interface 1503, and a bus 1504.
  • the memory 1501, the processor 1502, and the communication interface 1503 implement communication connections between each other through the bus 1504.
  • the memory 1501 may be a read only memory (Read Only Memory, ROM), a static storage device, a dynamic storage device, or a random access memory (Random Access Memory, RAM).
  • the memory 1501 may store a program. When the program stored in the memory 1501 is executed by the processor 1502, the processor 1502 and the communication interface 1503 are used to execute each step of the method described in the embodiment (1) or embodiment (2) of this application .
  • the processor 1502 may adopt a general-purpose central processing unit (Central Processing Unit, CPU), microprocessor, application specific integrated circuit (Application Specific Integrated Circuit, ASIC), graphics processing unit (graphics processing unit, GPU), or one or more
  • the integrated circuit is used to execute related programs to realize the functions required by the units in the positioning device 1400 of the embodiment of the present application, or to execute the method described in the embodiment (1) or the embodiment (2) of the method of the present application .
  • the processor 1502 may also be an integrated circuit chip with signal processing capability. In the implementation process, each step of the sample generation method of the present application can be completed by an integrated logic circuit of hardware in the processor 1502 or instructions in the form of software.
  • the aforementioned processor 1502 may also be a general-purpose processor, a digital signal processor (Digital Signal Processing, DSP), an application specific integrated circuit (ASIC), a ready-made programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic devices , Discrete gates or transistor logic devices, discrete hardware components.
  • DSP Digital Signal Processing
  • ASIC application specific integrated circuit
  • FPGA ready-made programmable gate array
  • FPGA Field Programmable Gate Array
  • the methods, steps, and logical block diagrams disclosed in the embodiments of the present application can be implemented or executed.
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like.
  • the steps of the method disclosed in the embodiments of the present application may be directly embodied as being executed and completed by a hardware decoding processor, or executed and completed by a combination of hardware and software modules in the decoding processor.
  • the software module can be located in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, or electrically erasable programmable memory, registers.
  • the storage medium is located in the memory 1501, and the processor 1502 reads the information in the memory 1501, and combines its hardware to complete the functions required by the units included in the positioning device 1400 of the embodiment of the present application, or execute the method embodiment (1) of the present application Or the method described in Example (2).
  • the communication interface 1503 uses a transceiving device such as but not limited to a transceiver to implement communication between the device 1500 and other devices or a communication network. For example, data such as point cloud data, first geometric features, second geometric features, and geometric feature maps can be acquired through the communication interface 1503.
  • the communication interface 1503 is also used to implement communication with other devices, such as a geometric feature map generating device, a map server, and a terminal.
  • the bus 1504 may include a path for transferring information between various components of the device 1500 (for example, the memory 1501, the processor 1502, and the communication interface 1503).
  • the vehicle When the device 1500 is installed on a vehicle, the vehicle may be the vehicle 100 shown in FIG. 3, and the device 1500 may also include a point cloud collection device 1505.
  • the positioning device 1500 can also execute the geometric feature extraction method in the first embodiment.
  • the point cloud acquisition device 1505 may be a laser radar (laser radar), a stereo camera (stereo camera), a time of flight camera (time of flight camera) and other devices that can acquire point cloud data.
  • the point cloud collection device 1505 is not a necessary device.
  • the acquiring unit 1401 in the apparatus 1400 may be equivalent to the communication interface 1503 in the apparatus 1500, and the extracting unit 1402 may be equivalent to the processor 1502.
  • the devices 1300 and 1500 shown in FIG. 13 and FIG. 15 only show a memory, a processor, a communication interface or a wireless communication module, in a specific implementation process, those skilled in the art should understand that the devices 1300 and 1500 1500 also includes other devices necessary for normal operation. At the same time, according to specific needs, those skilled in the art should understand that the devices 1300 and 1500 may also include hardware devices that implement other additional functions. In addition, those skilled in the art should understand that the apparatuses 1300 and 1500 may also include only the necessary devices for implementing the embodiments of the present application, and not necessarily all the devices shown in FIG. 13 or FIG. 15.
  • the devices 1300 and 1500 may also include an input/output device, and the input device may be a touch panel, a microphone, or other output devices.
  • the output device can be a display, an audio playback device, or other devices.
  • the devices 1300 and 1500 may also include various sensors, such as accelerometers, cameras, photosensitive sensors, fingerprint sensors, etc., which are not limited here.
  • the device 1300 may be equivalent to the positioning device in the embodiment corresponding to FIG. 2, such as the vehicle 160/terminal 180/positioning server 190, or the vehicle 100 in FIG. 3; the device 1500 may be equivalent to the positioning device in the embodiment corresponding to FIG. 2, such as The vehicle 160/terminal 180/positioning server 190 or the device 1500 may be equivalent to the geometric feature map generating device 140 in FIG. 2 or the vehicle 100 in FIG. 3.
  • the units and algorithm steps of the examples described in the embodiments disclosed in this document can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed by hardware or software depends on the specific application and design constraint conditions of the technical solution. Professionals and technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered beyond the scope of this application.
  • the disclosed system, device, and method may be implemented in other ways.
  • the device embodiments described above are merely illustrative, for example, the division of the units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components may be combined or It can be integrated into another system, or some features can be ignored or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium.
  • the technical solution of the present application essentially or the part that contributes to the existing technology or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including Several instructions are used to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disk or optical disk and other media that can store program code .
  • references described in this specification to "one embodiment” or “some embodiments”, etc. mean that one or more embodiments of the present application include a specific feature, structure, or characteristic described in conjunction with the embodiment. Therefore, the sentences “in one embodiment”, “in some embodiments”, “in some other embodiments”, “in some other embodiments”, etc. appearing in different places in this specification are not necessarily All refer to the same embodiment, but mean “one or more but not all embodiments” unless it is specifically emphasized otherwise.
  • the terms “including”, “including”, “having” and their variations all mean “including but not limited to”, unless otherwise specifically emphasized.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Multimedia (AREA)
  • Automation & Control Theory (AREA)
  • Navigation (AREA)

Abstract

一种定位方法及相关装置、设备，定位方法包括：定位设备通过从点云采集装置（126、1305、1505）中采集到的第一点云数据中提取出的N个第一几何特征对精确度低的第一位姿进行纠正，可得到精确度高的第二位姿，其采用数据量少的几何特征进行定位，大大减小了数据的运算量，使得车辆（100、160）定位的耗时少，定位的实时性好。

Description

定位方法、装置及系统 技术领域
本发明涉及人工智能领域的自动驾驶技术领域，尤其涉及一种定位方法、装置及系统。
背景技术
人工智能(Artificial Intelligence,AI)是利用数字计算机或者数字计算机控制的机器模拟、延伸和扩展人的智能，感知环境、获取知识并使用知识获得最佳结果的理论、方法、技术及应用系统。自动驾驶是人工智能领域的一种主流应用，自动驾驶技术依靠计算机视觉、雷达、监控装置和全球定位系统等协同合作，让机动车辆可以在不需要人类主动操作下，实现自动驾驶。自动驾驶的车辆使用各种计算系统来帮助将乘客从一个位置运输到另一位置。一些自动驾驶车辆可能要求来自操作者(诸如，领航员、驾驶员、或者乘客)的一些初始输入或者连续输入。自动驾驶车辆准许操作者从手动操作模式切换到自动驾驶模式或者介于两者之间的模式。由于自动驾驶技术无需人类来驾驶机动车辆，所以理论上能够有效避免人类的驾驶失误，减少交通事故的发生，且能够提高公路的运输效率。因此，自动驾驶技术越来越受到重视。
车辆的精准定位是实现自动驾驶的关键技术，目前基于高精度点云地图，车辆可以实现全局的高精度定位。具体的方法是：车辆通过点云采集装置采集的点云数据，将采集到的点云数据与点云地图进行点云配准(point cloud registration)，进而得到车辆的位姿。
然而，每公里点云数据的大小约4GB，若以点云数据进行定位，一方面，需要存储、传输和加载大量的点云数据到车辆的计算机系统；另一方面，车辆实时定位所需要运行的配准算法是基于大量的点云数据进行的运算，计算耗时长，很难满足车辆对定位的实时性要求，尤其是在车辆处于高速运动场景下，定位的实时性存在巨大挑战。
发明内容
本发明实施例所要解决的技术问题在于，提供一种定位方法、装置及系统，解决车辆定位过程计算量大，定位实时性差的技术问题。
第一方面，本发明实施例提供了一种定位方法，包括：定位设备获取车辆通过点云采集装置采集的第一点云数据；从该第一点云数据中提取N个第一几何特征，N为正整数；进而，根据N个第一几何特征对车辆的第一位姿进行调整以得到车辆的第二位姿，第二位姿的精确度高于第一位姿态的精确度。
应理解，第一位姿和第二位姿都是对车辆定位得到的位姿，第二位姿的精确度大于第一位姿的精确度。也可以说，第一位姿为车辆的预测位姿，第二位姿为车辆的实际位姿。
需要说明的是，该定位设备可以是上述车辆或车辆上的装置，也可以是终端，如手机、平板电脑等，也可以是定位芯片或定位装置，还可以是服务器或云端等。应理解，该终端或服务器可以与车辆进行通信连接，以获取到车辆观测到的第一点云数据。
上述方法提供了一种定位方法，通过从点云采集装置中采集到的第一点云数据中提取 出的N个第一几何特征对精确度低的第一位姿进行纠正，可得到精确度高的第二位姿，相对于现有技术中的点云数据，本申请实施例采用数据量少的几何特征进行定位，大大减小了数据的运算量，使得车辆定位的耗时少，定位的实时性好。
在本申请实施例一种可能的实现中，定位设备根据N个第一几何特征对车辆的第一位姿进行调整以得到该车辆的第二位姿的一种实现方式可以是：定位设备根据N个第一几何特征和几何特征地图中的N个第二几何特征对车辆的第一位姿进行调整以得到第二位姿，其中，几何特征地图为从点云地图的第二点云数据中提取的几何特征所形成的地图，N个第二几何特征为与N个第一几何特征匹配的几何特征。
在本申请实施例的一种可能的实现中，定位设备根据第一几何特征和几何特征地图中的第二几何特征对车辆的第一位姿进行调整以得到第二位姿的第一种实现方式可以是：
定位设备根据N个第一几何特征和几何特征地图中的N个第二几何特征确定几何特征之间的变换关系；进而，根据几何特征之间的变换关系对车辆的第一位姿进行调整以得到第二位姿。
上述方法，通过观测到的第一几何特征和几何特征地图中的第二几何特征之间的变换关系对精确度低的第一位姿进行纠正，可得到精确度高的第二位姿，相对于现有技术中的点云数据，本申请实施例采用数据量少的几何特征进行配准和定位，大大减小了数据的运算量，使得车辆定位的耗时少，定位的实时性好。
在本申请一种可能的实现中，定位设备根据N个第一几何特征和几何特征地图中的N个第二几何特征确定几何特征之间的变换关系可以包括但不限于如下两种实现方式
第一种实现方式：
定位设备通过第一变换量对N个第一几何特征进行变换以得到N个第三几何特征，第三几何特征与第一几何特征一一对应；进而，根据N个第三几何特征与N个第二几何特征之间的第一误差调节第一变换量；并且，在第一变换量的迭代次数满足停止迭代条件或第一误差满足停止迭代条件时，定位设备得到第一目标变换量，第一目标变换量为满足停止迭代条件时的第一变换量，第一目标变换量用于指示N个第一几何特征和N个第二几何特征之间的变换关系。
上述方法，将几何特征之间的变换关系的求解转换为N个第三几何特征与N个第二几何特征之间的第一误差的最小化，进而，通过迭代确定第一误差最小时第一变换量为N个第一几何特征和N个第二几何特征之间的变换关系，使得得到的变换关系更加准确。
可选地，定位设备可以根据第一目标函数确定第一误差，其中，第一目标函数可以为：
Figure PCTCN2019105810-appb-000001
ε为第一误差；第一变换量包括旋转R和平移t；w i为第一几何特征V i的权重；U i为第一几何特征V i对应的第二几何特征；i为N个第一几何特征中第一几何特征的索引，i为正整数，i≤N。
第二种实现方式：
定位设备通过第二变换量对N个第二几何特征进行变换以得到N个第四几何特征，第 四几何特征与第二几何特征一一对应；进而，根据N个第四几何特征与N个第一几何特征之间的第二误差调节第二变换量；并且，在第二变换量的迭代次数停止迭代条件或第二误差满足停止迭代条件时，定位设备得到第二目标变换量，第二目标变换量为满足停止迭代条件时第二变换量的逆矩阵，第二目标变换量用于指示N个第一几何特征和N个第二几何特征之间的变换关系。
上述方法，将几何特征之间的变换关系的求解转换为N个第四几何特征与N个第一几何特征之间的第二误差的最小化，进而，通过迭代确定第二误差最小时第二变换量为N个第一几何特征和N个第二几何特征之间的变换关系，使得得到的变换关系更加准确。
可选地，定位设备可以根据第二目标函数确定第二误差，其中，第二目标函数可以为：
Figure PCTCN2019105810-appb-000002
ε为第二误差；第二变换量包括旋转R′和平移t′；w i为第一几何特征V i的权重；v i为第一几何特征V i中的向量，u i为第一几何特征V i对应的第二几何特征U i中的向量；N为第一几何特征的个数，i为N个第一几何特征中第一几何特征的索引，i为正整数，i≤N。
可选地，在上述第一目标函数或第二目标函数中，第一几何特征V i的权重w i与第一几何特征V i所属的对象相对于车辆的距离负相关。即，距离车辆越近的对象对应的第一几何特征对变换关系的贡献越大，由于与车辆越近的第一几何特征，该提取到的第一几何特征的误差越小，使得定位设备在确定变换关系使可以更多地基于准确度搞的第一几何特征，在提高变换关系的精确度，进而提高定位的精确度。
在本申请实施例的一种可能的实现中，定位设备根据第一几何特征和几何特征地图中的第二几何特征对车辆的第一位姿进行调整以得到第二位姿的第二种实现方式可以是：
定位设备根据第一位姿，对车辆的位姿进行估计以得到多组估计位姿；进而，根据N个第一几何特征和几何特征地图中的N个第二几何特征确定多组估计位姿的评分；从而，定位设备根据多组估计位姿中每一组估计位姿的评分确定车辆的第二位姿，其中，第一组估计位姿的评分用于指示第一组估计位姿与第二位姿的接近程度，第一组估计位姿是多组估计位姿中任意一个估计位姿。
应理解，第一位姿、估计位姿和第二位姿都是对车辆定位得到的位姿，第二位姿的精确度大于第一位姿的精确度。也可以说，第一位姿为车辆的预测位姿，第二位姿为车辆的实际位姿。多组估计位姿是分布于第一位姿周围的位姿，评分满足条件的估计位姿，如评分最大的估计位姿即为第二位姿。
上述方法提供了一种车辆定位方法，通过观测到的第一几何特征和几何特征地图中的第二几何特征对估计位姿进行评分，确定评分最高的估计位姿作为车辆的实际位姿，相对于现有技术中的点云数据，本申请实施例采用数据量少的几何特征进行定位，大大减小了数据的运算量，使得车辆定位的耗时少，定位的实时性好。
在本申请一种可能的实现中，定位设备根据N个第一几何特征和几何特征地图中的N个第二几何特征确定多组估计位姿的评分，可以包括但不限于如下三种实现方式：
实现方式A：
定位设备根据每一组估计位姿和N个第二几何特征，确定每一组估计位姿对应的第一参数的估计值；根据第一位姿和N个第一几何特征，确定第一参数的观测值；进而，定位设备根据每一组估计位姿对应的第一参数的估计值与第一参数的观测值之间的误差确定每一组估计位姿的评分。
可选地，第一参数为距离、方位角和高度角中的至少一种；每一组估计位姿对应的第一参数的估计值为N个第二几何特征分别相对于每一组估计位姿下的车辆的第一参数；第一参数的观测值为N个第一几何特征分别相对于在第一位姿下的车辆的第一参数。
上述方法通过将估计位姿的评分转换为估计位姿下第一参数的估计值与第一参数的实际观测值之间是误差，使得估计位姿评估过程更加简单，进一步地降低定位耗时。
实现方式B：
定位设备通过每一组估计位姿与第一位姿之间的变换关系对N个第二几何特征分别进行变换得到的每一组估计位姿对应的N个第五几何特征，第二几何特征与第五几何特征一一对应；进而，定位设备根据每一组估计位姿对应的N个第五几何特征与N个第一几何特征之间的误差确定每一组估计位姿的评分。
上述方法，将估计位姿的评分转换为估计位姿下第一参数的估计值与实际的观测值之间的误差，使得估计位姿评估过程更加简单，进一步地降低定位耗时。
实现方式C：
定位设备通过每一组估计位姿与第一位姿之间的变换关系对N个第一几何特征分别进行变换得到的每一组估计位姿对应的N个第六几何特征，第一几何特征与第六几何特征一一对应；进而，定位设备根据每一组估计位姿对应的N个第六几何特征与N个第二几何特征之间的误差确定每一组估计位姿的评分。
在本申请一种可能的实现中，在定位设备根据N个第一几何特征对车辆的第一位姿进行调整以得到车辆的第二位姿之前，定位设备还可以获取车辆的第一位姿，定位设备获取车辆的第一位姿可以包括但不限于如下两种实现：
实现(一)：定位设备根据上一时刻的第二位姿确定车辆在当前时刻的预测位姿，当前时刻的预测位姿为定位设备获取的车辆的第一位姿，上一时刻为当前时刻之前的时刻。例如，定位设备可以将车辆的上一时刻的准确的位姿(即上一时刻的第二位姿)、车辆在上一时刻的控制参数输入到车辆的运动学方程，预测当前时刻的第一位姿。
上述方法基于准确度高的上一时刻的位姿进行估计，估计得到的当前时刻的第一位姿与车辆的实际位姿更加接近，减少运算的迭代次数，进一步地，提高车辆第二位姿的计算效率和车辆定位的响应速度。
实现(二)：定位设备根据定位系统确定车辆的第一位置以及根据惯性传感器确定车辆的第一姿态，第一位姿包括第一位置和第一姿态。
在本申请实施例的一种可能的实现中，定位设备根据第一几何特征和几何特征地图中的第二几何特征对车辆的第一位姿进行调整以得到第二位姿的第三种实现方式可以是：
定位设备根据上一时刻的第二位姿和上一时刻时车辆的控制参数，确定车辆在当前时刻的预测位姿，该上一时刻为当前时刻之前的时刻；进而，定位设备通过第二参数的观测值与第二参数的预测值之间的误差更新车辆的预测位姿，得到车辆的第二位姿，其中，第 二参数的观测值是基于车辆在第一位姿下观测到的第一几何特征确定的，第二参数的预测值是基于预测位姿和几何特征地图中的第二几何特征确定的。
应理解，上述第一位姿可以是当前时刻的预测位姿，也可以是通过其他方法定位到的车辆当前时刻的位姿，比如，定位设备根据定位系统确定车辆的第一位置以及根据惯性传感器确定车辆的第一姿态，第一位姿包括第一位置和第一姿态。
还应理解，第一位姿、当前时刻的预测位姿和当前时刻的第二位姿都是对当前时刻下车辆定位得到的位姿。第二位姿的精确度大于预测位姿的精确度，也大于第一位姿的精确度。也可以说，第一位姿、当前时刻的预测位姿为当前时刻下对车辆的预测得到的位姿，第二位姿为当前时刻下车辆的实际位姿。
上述方法，通过观测到的第一几何特征确定的第二参数的观测值和基于几何特征地图中的第二几何特征确定的第二参数的预测值之间的误差对当前时刻的预测位姿进行更新，得到车辆的实际位姿，相对于现有技术中的点云数据，本申请实施例采用数据量少的几何特征进行定位，大大减小了数据的运算量，使得车辆定位的耗时少，定位的实时性好。
在本申请一种可能的实现中，定位设备基于第一位姿、预测位姿、N个第一几何特征和N个第二几何特征确定第二参数的观测值与第二参数的预测值之间的误差的一种实现方式可以是：定位设备根据预测位姿和N个第二几何特征确定第二参数的预测值；根据第一位姿和N个第一几何特征确定第二参数的观测值；进而，根据第二参数的观测值和第二参数的预测值确定第二参数的观测值与第二参数的预测值之间的误差。
上述方法采用了卡尔曼滤波的方法来确定第二位姿，更新次数少，减少运算过程，定位快。
可选地，第二参数为距离、方位角和高度角中的至少一种；第二参数的预测值为N个第二几何特征分别相对于预测位姿下的车辆的第二参数；第二参数的观测值为N个第一几何特征分别相对于在第一位姿下的车辆的第二参数。
上述方法通过预测位姿下第二参数的预测值与第二参数的实际观测值之间的误差来衡量该预测位姿与车辆的实际位姿之间的误差，更新预测位姿使得第二参数的预测值与第二参数的实际观测值之间的误差最小，即得到车辆的实际位姿，即第二位姿。上述进一步地降低定位耗时。
需要说明的是，该定位设备可以是上述车辆或车辆上的装置，也可以是终端，如手机、平板电脑等，也可以是定位装置、定位芯片，还可以是服务器或云端等，该终端或服务器可以与车辆进行通信连接，以获取到车辆观测到的第一几何特征。
在本申请实施例的一种可能的实现中，第一点云数据为在由第一位姿确定的空间中表示的车辆观测到的对象的表面上的点的信息，N个第一几何特征中每个第一几何特征用于指示在由第一位姿确定的空间中该车辆观测到的一个对象的几何特征。
上述方法，定位设备从数据量大的点云数据中提取出对象的几何特征，即N个第一几何特征，通过数据量少的几何特征进行配准和定位，大大减小了数据的运算量，提高定位的响应速度。
在本申请一种可能的实现中，该方法还包括：定位设备在几何特征地图中查找与N个第一几何特征相匹配的N个第二几何特征。
在本申请一种可能的实现中，定位设备在几何特征地图中查找与N个第一几何特征相匹配的N个第二几何特征的可以包括的一种实现方式可以是：定位设备从几何特征地图的第一区域中查找与N个第一几何特征相匹配的N个第二几何特征，该第一区域为基于第一位姿确定的区域，第一区域不小于车辆的点云采集装置的扫描范围。
在该实现方式中，定位设备根据第一位姿确定第一区域，缩小查找范围，提高计算效率。
可选地，定位设备从几何特征地图的第一区域中查找与N个第一几何特征相匹配的N个第二几何特征可以包括但不限于如下两种实现方式可以是：
实现方式一：定位设备针对第一几何特征V i，确定第一几何特征V i与第一区域中每一个几何特征的偏差；进而，将第一区域中的几何特征中与第一几何特征V i的偏差最小的几何特征作为与第一几何特征V i相匹配的第二几何特征U i，其中，i＝1,2，…，N。
上述方法通过计算计算两个几何特征之间的偏差进行匹配，提高匹配准确度。
实现方式二：定位设备针对第一几何特征V i，从第一区域中的几何特征中选择与第一几何特征V i的属性相匹配的几何特征；进而，将属性相匹配的几何特征作为与第一几何特征V i相匹配的第二几何特征U i，其中，i＝1,2，…，N。
上述方法，通过该属性来匹配，更加简单，提高匹配效率。
应理解，当与第一几何特征V i属的性相匹配的几何特征为多个时，定位设备可以基于实现方式一中的方法，确定与第一几何特征V i相匹配的第二几何特征U i为上述与第一几何特征V i属的性相匹配的多个几何特征中与第一几何特征V i的偏差最小的几何特征。
在上述实现方式一和实现方式二中，在几何特征进行匹配过程中，可以仅考虑第一几何特征中的向量或位置坐标，也可以同时考虑两者，此处不作限定。
在本申请一种可能的实现中，定位设备从第一点云数据中提取N个第一几何特征的一种实现方式可以是：定位设备识别第一点云数据中的N个对象；进而，定位设备基于该N个对象中每个对象的点云数据确定每个对象的第一几何特征。
可选地，第一对象为多个对象中的任意一个对象，以基于第一对象的点云数据确定第一对象的第一几何特征为例来说明基于N个对象中每个对象的点云数据确定每个对象的第一几何特征：
若第一对象的几何形状为直线，对第一对象的点云数据进行直线拟合以得到第一对象的第一几何特征，第一对象的第一几何特征为拟合得到的直线的几何特征；
若第一对象的几何形状为曲线，对第一对象的点云数据进行曲线拟合以得到第一对象的第一几何特征，第一对象的第一几何特征为拟合得到的曲线的几何特征；
若第一对象的几何形状为平面，对第一对象的点云数据进行平面拟合以得到第一对象的第一几何特征，第一对象的第一几何特征为拟合得到的平面的几何特征；
若第一对象的几何形状为曲面，对第一对象的点云数据进行曲面拟合以得到第一对象的第一几何特征，第一对象的第一几何特征为曲面得到的曲线的几何特征。
上述方法，提供一种从点云数据中提取出第一几何特征的方法。首先识别对象的几何形状，进而通过识别到的几何形状对应的拟合方法进行拟合，将拟合得到的直线/曲线/平面/曲面的几何特征作为该对象的第一几何特征，可以提高提取的第一几何特征的准确度。
第二方面，本申请实施例还提供了一种几何特征提取方法，包括：执行设备获取待处理的点云数据；从该待处理的点云数据中提取至少一个几何特征；其中，该至少一个几何特征用于车辆的定位。
在本申请实施例的一种实现中，执行设备可以是几何特征地图生成设备，待处理的点云数据可以是点云地图中的第二点云数据，此时，从第二点云数据中提取出的几何特征形成几何特征地图，该几何特征地图用于车辆的通过几何特征进行定位。
在本申请实施例的一种实现中，执行设备可以是定位设备，待处理的点云数据也可以是车辆通过点云采集装置采集到的第一点云数据，此时，从第一点云数据中提取的N个第一几何特征用于对该车辆的第一位姿进行调整以得到该车辆的第二位姿。
在本申请实施例的一种实现中，执行设备从待处理的点云数据中提取至少一个几何特征的一种实现方式可以是：
执行设备识别待处理的云数据中的至少一个对象；进而，定位设备基于该至少一个对象中每个对象的点云数据确定每个对象的几何特征。
可选地，第二对象是至少一个对象中任意一个对象，以基于第二对象的点云数据确定第二对象的几何特征为例来说明基于至少一个对象中每个对象的点云数据确定每个对象的几何特征，第二对象的点云数据确定第二对象的几何特征的方法可以是：
若第二对象的几何形状为直线，执行设备对第二对象的点云数据进行直线拟合以得到第二对象的几何特征，第二对象的几何特征为拟合得到的直线的几何特征；
若第二对象的几何形状为曲线，执行设备对第二对象的点云数据进行曲线拟合以得到第二对象的几何特征，第二对象的几何特征为拟合得到的曲线的几何特征；
若第二对象的几何形状为平面，执行设备对第二对象的点云数据进行平面拟合以得到第二对象的几何特征，第二对象的几何特征为拟合得到的平面的几何特征；
若第二对象的几何形状为曲面，执行设备对第二对象的点云数据进行曲面拟合以得到第二对象的几何特征，第二对象的几何特征为曲面得到的曲线的几何特征。
第三方面，本申请实施例还提供了一种定位装置，包括：
第一获取单元，用于车辆通过点云采集装置采集的第一点云数据；
特征提取单元，用于从该第一点云数据中提取N个第一几何特征，N为正整数；
调整单元，用于根据该N个第一几何特征对该车辆的第一位姿进行调整以得到该车辆的第二位姿，该第二位姿的精确度高于该第一位姿的精确度。
在本申请一种可能的实现中，该装置还包括：
第二获取单元，用于在该调整单元根据该N个第一几何特征对该车辆的第一位姿进行调整以得到该车辆的第二位姿之前，获取该车辆的第一位姿。
在本申请一种可能的实现中，该装置还包括：
匹配单元，用于在该几何特征地图中查找与该N个第一几何特征相匹配的N个第二几何特征。
需要说明的是，上述第三方面的定位装置还包括其他用于实现第一方面所述的定位方 法中的单元，上述定位装置的各个单元或其他单元的具体实现以参见上述第一方面中相关描述，此处不再赘述。
第四方面，本申请实施例还提供了一种定位装置，包括：处理器和存储器，该存储器用于存储程序，该处理器执行该存储器存储的程序，当该存储器存储的程序被执行时，可实现如权利要求第一方面或第一方面的任一种实现所述的方法。
上述定位装置还包括其他用于实现第一方面所述的定位方法中的器件或模块，上述定位装置的各个器件或其他器件的具体实现以参见上述第一方面中相关描述，此处不再赘述。
第五方面，本申请实施例还提供了一种几何特征的提取装置，包括：
获取单元，用于获取待处理的点云数据；
提取单元，用于从该待处理的点云数据中提取至少一个几何特征；其中，该至少一个几何特征用于车辆的定位。
在本申请实施例的一种实现中，几何特征的提取装置可以是几何特征地图生成设备，待处理的点云数据可以是点云地图中的第二点云数据，此时，从第二点云数据中提取出的几何特征形成几何特征地图，该几何特征地图用于车辆的通过几何特征进行定位。
在本申请实施例的一种实现中，几何特征的提取装置可以是定位设备，待处理的点云数据也可以是车辆通过点云采集装置采集到的第一点云数据，此时，从第一点云数据中提取的N个第一几何特征用于对该车辆的第一位姿进行调整以得到该车辆的第二位姿。
需要说明的是，上述第五方面所述的几何特征的提取装置还包括其他用于实现第二方面所述的几何特征的提取方法中的单元，上述定位装置的各个单元或其他单元的具体实现以参见上述第二方面中相关描述，此处不再赘述。
第六方面，本申请实施例还提供了一种几何特征的提取装置，包括：处理器和存储器，该存储器用于存储程序，该处理器执行所述存储器存储的程序，当该存储器存储的程序被执行时，可实现如第二方面或第二方面的任一种实现所述的方法。
上述定位装置还包括其他用于实现第二方面所述的几何特征的提取方法中的器件或模块，上述定位装置的各个器件或其他器件的具体实现以参见上述第二方面中相关描述，此处不再赘述。
第七方面，本申请实施例还提供了一种车辆，包括：点云采集装置、处理器和存储器，该处理器通过总线连接到该点云采集装置，该点云采集装置用于采集点云数据，该存储器用于存储程序，该处理器执行该存储器存储的程序，当该存储器存储的程序被执行时，可实现如第一方面或第一方面的任一种实现所述的方法。
第八方面，本申请实施例还提供了一种车辆，包括：点云采集装置、处理器和存储器，该处理器通过总线连接到该点云采集装置，该点云采集装置用于采集点云数据，该存储器用于存储程序，该处理器执行该存储器存储的程序，当该存储器存储的程序被执行时，可实现如第二方面或第二方面的任一种实现所述的方法。
第九方面，本申请实施例还提供了一种计算机可读存储介质，该计算机存储介质存储有计算机程序，该计算机程序包括程序指令，该程序指令当被处理器执行时，可实现如第一方面中的方法。
第十方面，本申请实施例还提供了一种计算机可读存储介质，该计算机存储介质存储有计算机程序，该计算机程序包括程序指令，该程序指令当被处理器执行时，可实现如第二方面中的方法。
第十一方面，本申请实施例还提供了提供一种包含指令的计算机程序产品，当该计算机程序产品在计算机上运行时，使得计算机执行上述第一方面中的方法。
第十二方面，本申请实施例还提供了提供一种包含指令的计算机程序产品，当该计算机程序产品在计算机上运行时，使得计算机执行上述第二方面中的方法。
第十三方面，提供一种定位芯片，该芯片包括处理器与数据接口，该处理器通过该数据接口读取存储器上存储的指令，执行第一方面中的方法。
可选地，作为一种实现方式，该芯片还可以包括存储器，该存储器中存储有指令，该处理器用于执行该存储器上存储的指令，当该指令被执行时，该处理器用于执行第一方面中的方法。
第十四方面，提供一种芯片，该芯片包括处理器与数据接口，该处理器通过该数据接口读取存储器上存储的指令，执行第二方面中的方法。
可选地，作为一种实现方式，该芯片还可以包括存储器，该存储器中存储有指令，该处理器用于执行该存储器上存储的指令，当该指令被执行时，该处理器用于执行第二方面中的方法。
第十五方面，提供一种电子设备，该电子设备包括上述第三方面或第四方面中的任意一个方面中的定位装置。
第十六方面，提供一种电子设备，该电子设备包括上述第五方面或第六方面中的任意一个方面中的定位装置。
第十七方面，提供一种定位方法，该方法可以包括：定位设备接收N个第一几何特征，进而，定位设备根据N个第一几何特征对车辆的第一位姿进行调整以得到车辆的第二位姿，其中，N个第一几何特征是从第一点云数据中提取得到的，第一点云数据是车辆通过点云采集装置采集的点云数据，第二位姿的精确度高于第一位姿的精确度。
可选地，定位设备根据N个第一几何特征对车辆的第一位姿进行调整以得到车辆的第二位姿的具体实现可以参见上述第一方面中相关描述，此处不再赘述。
可选地，定位设备获取N个第一几何特征的方式可以是：定位设备接收终端/车辆发送的N个第一几何特征，该N个第一几何特征是终端/车辆从车辆通过点云采集装置采集的第一点云数据中提取得到的。
需要说明的是，终端/车辆从车辆通过点云采集装置采集的第一点云数据中提取出N个第一几何特征的具体实现可以参见上述第一方面中定位设备从第一点云数据中提取出N个第一几何特征的具体实现，此处不再赘述。
需要说明的是，定位设备根据N个第一几何特征对车辆的第一位姿进行调整以得到该车辆的第二位姿的具体实现可以参见上述第一方面中相关描述，此处不再赘述。
第十八方面，本申请实施例还提供了一种定位装置，包括：
接收单元，用于接收N个第一几何特征，其中，该N个第一几何特征是从第一点云数据中提取得到的，该第一点云数据是车辆通过点云采集装置采集的点云数据；
调整单元，用于根据该N个第一几何特征对该车辆的第一位姿进行调整以得到该车辆的第二位姿，该第二位姿的精确度高于该第一位姿的精确度。
在本申请一种可能的实现中，该装置还包括：
获取单元，用于在该调整单元根据该N个第一几何特征对该车辆的第一位姿进行调整以得到该车辆的第二位姿之前，获取该车辆的第一位姿。
在本申请一种可能的实现中，该装置还包括：
匹配单元，用于在该几何特征地图中查找与该N个第一几何特征相匹配的N个第二几何特征。
需要说明的是，上述第十八方面所述的定位装置还包括其他用于实现第十七方面所述的定位方法中的单元，上述定位装置的各个单元或其他单元的具体实现以参见上述第十七方面中相关描述，此处不再赘述。
第十九方面，本申请实施例还提供了一种定位装置，包括：处理器和存储器，该存储器用于存储程序，该处理器执行该存储器存储的程序，当该存储器存储的程序被执行时，可实现如第十七方面或第十七方面的任一种实现所述的方法。
上述定位装置还包括其他用于实现第十七方面所述的定位方法中的器件或模块，上述定位装置的各个器件或其他器件的具体实现以参见上述第十七方面中相关描述，此处不再赘述。
第二十方面，本申请实施例还提供了一种计算机可读存储介质，该计算机存储介质存储有计算机程序，该计算机程序包括程序指令，该程序指令当被处理器执行时，可实现如第十七方面中的方法。
第二十一方面，本申请实施例还提供了提供一种包含指令的计算机程序产品，当该计算机程序产品在计算机上运行时，使得计算机执行上述第十七方面中的方法。
第二十二方面，提供一种定位芯片，该芯片包括处理器与数据接口，该处理器通过该数据接口读取存储器上存储的指令，执行第十七方面中的方法。
可选地，作为一种实现方式，该芯片还可以包括存储器，该存储器中存储有指令，该处理器用于执行该存储器上存储的指令，当该指令被执行时，该处理器用于执行第十七方面中的方法。
第二十三方面，提供一种电子设备，该电子设备包括上述第十八方面或第十九方面中的任意一个方面中的定位装置。
第二十四方面，提供一种定位方法，该方法可以包括：车辆通过点云采集装置采集的第一点云数据；进而，该车辆从该第一点云数据中提取N个第一几何特征获取N个第一几何特征，进而，该车辆将该N个第一几何特征发送至定位设备，以使该定位设备在接收到N个第一几何特征后，该定位设备根据N个第一几何特征对该车辆的第一位姿进行调整以得到该车辆的第二位姿，其中，第二位姿的精确度高于第一位姿的精确度。
需要说明的是，定位设备根据N个第一几何特征对车辆的第一位姿进行调整以得到车辆的第二位姿的具体实现可以参见上述第一方面中相关描述，此处不再赘述。
需要说明的是，车辆从第一点云数据中提取N个第一几何特征的实现方式可以参照上述第一方面中定位设备从第一点云数据中提取N个第一几何特征的具体实现，此处不再赘 述。
第二十五方面，本申请实施例还提供了一种定位装置，包括：
采集单元，用于通过点云采集装置采集的第一点云数据；
提取单元，用于从该第一点云数据中提取N个第一几何特征获取N个第一几何特征；
发送单元，用于将该N个第一几何特征发送至定位设备，以使该定位设备在接收到N个第一几何特征后，该定位设备根据N个第一几何特征对该车辆的第一位姿进行调整以得到该车辆的第二位姿，其中，第二位姿的精确度高于第一位姿的精确度。
在本申请一种可能的实现中，该装置还包括：
获取单元，用于在该调整单元根据该N个第一几何特征对该车辆的第一位姿进行调整以得到该车辆的第二位姿之前，获取该车辆的第一位姿。
在本申请一种可能的实现中，该装置还包括：
匹配单元，用于在该几何特征地图中查找与该N个第一几何特征相匹配的N个第二几何特征。
需要说明的是，上述第二十五方面所述的定位装置还包括其他用于实现第二十四方面所述的定位方法中的单元，上述定位装置的各个单元或其他单元的具体实现以参见上述第二十四方面中相关描述，此处不再赘述。
第二十六方面，本申请实施例还提供了一种定位装置，包括：处理器、存储器和通信接口，该存储器用于存储程序，该处理器执行该存储器存储的程序，当该存储器存储的程序被执行时，可实现如第二十四方面或第二十四方面的任一种实现所述的方法。
上述定位装置还包括其他用于实现第二十四方面所述的定位方法中的器件或模块，上述定位装置的各个器件或其他器件的具体实现以参见上述第二十四方面中相关描述，此处不再赘述。
第二十七方面，本申请实施例还提供了一种计算机可读存储介质，该计算机存储介质存储有计算机程序，该计算机程序包括程序指令，该程序指令当被处理器执行时，可实现如第二十四方面中的方法。
第二十八方面，本申请实施例还提供了提供一种包含指令的计算机程序产品，当该计算机程序产品在计算机上运行时，使得计算机执行上述第二十四方面中的方法。
第二十九方面，提供一种芯片，该芯片包括处理器与数据接口，该处理器通过该数据接口读取存储器上存储的指令，执行第二十四方面中的方法。
可选地，作为一种实现方式，该芯片还可以包括存储器，该存储器中存储有指令，该处理器用于执行该存储器上存储的指令，当该指令被执行时，该处理器用于执行第二十四方面的方法。
第三十方面，提供一种电子设备，该电子设备包括上述第二十五方面或第二十六方面中的任意一个方面中的定位装置。
第三十一方面，本申请实施例还提供了一种车辆，包括：点云采集装置、处理器和存储器，该处理器通过总线连接到该点云采集装置，该点云采集装置用于采集点云数据，该存储器用于存储程序，该处理器执行该存储器存储的程序，当该存储器存储的程序被执行时，可实现如第二十四方面或第二十四方面的任一种实现所述的方法。
附图说明
为了更清楚地说明本发明实施例或背景技术中的技术方案，下面将对本发明实施例或背景技术中所需要使用的附图进行说明。
图1A是本发明实施例提供的一种道路场景示意性说明图；
图1B是本发明实施例提供的一种对应于图1A的几何特征示意性说明图；
图2是本发明实施例提供的一种系统的框架示意图；
图3是本发明实施例提供的一种车辆的功能框图；
图4A是本发明实施例提供的一种几何特征提取方法的流程示意图；
图4B是本发明实施例提供的一种提取路面和路沿的几何特征的原理示意图；
图4C是本发明实施例提供的另一种提取路面和路沿的几何特征的原理示意图；
图4D是本发明实施例提供的一种提取道路指示线的几何特征的几何特征的原理示意图；
图4E是本发明实施例提供的一种树木的点云数据的分布示意图；
图4F是本发明实施例提供的一种广告牌的几何特征的示意性说明图；
图4G是本发明实施例提供的一种路灯杆的几何特征的示意性说明图；
图4H是本发明实施例提供的一种指示牌上点的距离和角度的示意性说明图；
图4I是本发明实施例提供的一种建筑物的点云数据的分布示意图；
图5是本发明实施例提供的一种的几何地图的生成方法的流程示意图；
图6是本发明实施例提供的一种定位原理示意图；
图7是本申请实施例提供的一种定位方法的流程示意图；
图8A是本申请实施例提供的一种第一区域的示意性说明图；
图8B是本申请实施例提供的另一种第一区域的示意性说明图；
图8C是本申请实施例提供的又一种第一区域的示意性说明图；
图9A是本申请实施例提供的另一种定位方法的流程示意图；
图9B是本申请实施例提供的一种第一定位方式的一种流程示意图；
图9C是本申请实施例提供的一种第二定位方式的一种流程示意图；
图10A是本申请实施例提供的又一种定位方法的流程示意图；
图10B是本申请实施例提供的一种第三定位方式的流程示意图；
图10C是本申请实施例提供的一种第三种定位方式的示意性说明图；
图10D是本申请实施例提供的一种向量的高度角和方位角的示意说明图；
图11A是本申请实施例提供的又一种定位方法的流程示意图；
图11B是本申请实施例提供的一种第四定位方式的流程示意图；；
图11C是本申请实施例提供的一种第五定位方式的一种流程示意图；
图11D是本申请实施例提供的一种第五定位方式的一种流程示意图；
图12是本申请实施例提供的一种定位装置的结构示意图；
图13是本申请实施例提供的一种定位装置的硬件结构示意图；
图14是本申请实施例提供的一种几何特征提取装置的结构示意图；
图15是本申请实施例提供的另一种几何特征提取装置的硬件结构示意图。
具体实施方式
在介绍本申请具体实施方式之前，首先介绍描述本申请具体实施方式中用到的缩略语、中英文对照以及关键术语定义。
(1)点云数据(point cloud data)
点云数据(本申请中也称为“点云”)的本质为三维空间中点的集合，通常用空间中三维坐标来表示，是物体的表面的信息转化得到的点的信息。除位置坐标外，一个点的数据还可以包括RGB颜色、灰度值、深度、强度信息、分割结果等。其中，强度信息为激光雷达接收到的回波强度，该强度信息与目标的表面材质、粗糙度、入射角，激光雷达发射波的能量、波长等有关；分割结果可以是该点所在的物体的标识、属性、该点在所在物体上所处的位置或区域等。
例如，P i＝{x i,y i,z i,R i,G i,B i,Q i,……}表示空间中的一个点，其中，(x i,y i,z i)为点P i在坐标系OXYZ下的空间坐标，其中，OX、OY和OZ相互垂直为相互垂直的坐标轴；(R i,G i,B i)分别为点P i处红(R)、绿(G)、蓝(B)三个颜色的亮度；Q i为点P i处对激光的反射强度或反射率。此时，C＝{P 1,P 2,P 3,…，P n}表示一组点云数据。
通常点云数据是由点云采集装置例如激光雷达(laser radar)，立体摄像头(stereo camera)，越渡时间(timeof flight,TOF)摄像头采集到的。激光雷达完成一次扫描，即可得到一帧点云数据。
(2)点云地图(point cloud map)
配置有激光雷达传感器、双摄像头等点云采集装置的车辆可以采集各个道路的点云数据，进而，通过地图构建的方法，例如同步定位与地图构建(simultaneous localization and mapping，SLAM)算法，将点云数据逐帧累加构建成点云地图。应理解，在点云地图通常是在世界坐标系(英文：world coordinate system，也称全局坐标系)下的点云数据。
(3)点云配准(point cloud registration)
点云配准是将两组来自不同坐标系(也称点集空间)的点云数据经过旋转、平移等刚性变换统一到指定坐标系下的过程。也就是说，进行配准的两个点云，它们彼此之间可以通过旋转平移等变换后完全重合，因此，点云配准就是找到两个点云之间的坐标的变换关系。
对于定位不够准确的车辆来说，车辆的点云采集装置采集到的点云数据是在当前不准确的位置(本申请中也称“预测位置”)和不准确的姿态(本申请中也称为“预测姿态”)所确定的坐标系下的点云数据。通过该坐标系与点云地图中点云数据所采用的坐标系(即世界坐标系)之间的变换关系，可以得到车辆准确的位置(本申请中也称“实际位置”)和准确的姿态(本申请中也称“实际姿态”)。其中，实际位置是指车辆在世界坐标系下的位置，实际姿态是指车辆准确的姿态。应理解，本申请中“位姿(pose)”是指“位置(position)”和“姿态(attitude)”，同理，预测位姿是指预测位置和预测姿态，实际位姿是指实际位置和实际姿态。
首先，可以找到点云地图中与车辆采集到的点云数据相匹配的点云数据；进而，根据 车辆采集到的点云数据和点云地图中与车辆采集到的点云数据相匹配的点云数据，通过匹配算法可以计算出上述两组点云数据之间的坐标的变换关系。进一步地，将预测位姿S predict通过计算得到的变换关系T进行变换，得到车辆的实际位置S real。可通过数学关系表达为：S real＝TS predict，其中，
Figure PCTCN2019105810-appb-000003
表示两个点云之间存在旋转R和平移t的变换。
(4)位姿
位姿是指车辆在某一坐标系下的位置和姿态。其中，位置是指车辆在坐标系下的位置坐标；姿态是指车辆分别绕坐标系中x轴、y轴和z轴的旋转角度，分别为俯仰角(pitch)、偏航角(yaw)、翻滚角(roll)，共6个自由度(degree of freedom，DoF)。
由于，车辆一般运行在地面道路上，z轴(即垂直于地平面的坐标轴)坐标一般不需要进行自由变换，因此，位姿也可以选为3DoF来表达，即x轴坐标、y轴坐标和偏航角。进而，可以减少点云数据的维度，降低点云数据的数据量。应理解，上述位姿的表达方式仅为示例性说明，本申请中位姿也可以4DoF(即x轴坐标、y轴坐标、z轴坐标和偏航角)或其他个数的自由度来表示，本申请对此不作限定。
(5)几何特征
目前，每公里点云数据的大小约GB级或十几GB级，若将以点云数据用于定位，一方面，需要存储、传输和加载大量的点云数据到车辆的计算机系统；另一方面，车辆实时定位所需要运行的配准算法是基于大量的点云数据进行的运算，计算耗时长，很难满足实时性要求，尤其是在车辆处于高速运动场景下，算法的实时性存在巨大挑战。
然而，本申请提出的定位方法，不是基于点云数据的进行配准，而是基于从点云数据中提取的几何特征来进行配准。
几何特征是从点云数据中提取得到的，点云数据描述了环境的中物体的表面的采样点的信息。本申请实施例中环境中物体的几何形状主要包括线和面，其中，面可以包括平面和曲面，线可以包括直线和曲线。一个物体可以包括但不限于多个几何特征，一个几何特征所描述的对象可以是物体上的几何形状为线或面的部分。几何形状为平面的对象的几何特征可以包括平面的法向量、平面的位置等；曲面的几何特征可以包括多个法向量、多个主方向、曲面上多个坐标点、多项式曲面的系数等中的至少一种；直线的几何特征可以包括方向向量、直线的位置；曲线的几何特征可以包括曲线的多个切向量、多个法向量、多个坐标点、多项式曲线的系数等中的至少一种。
上述几何特征中的各个向量即为该几何特征所描述对象的方向向量和/或位置，方向向量可以通过矢量来表示，其中，该矢量可以是单位矢量。例如，路面的几何特征，即为路面所在平面的法向量，可以表示为(0,0,1)。需要说明的是，上述(0,0,1)仅为示例性说明。又例如，路沿的方向向量为(0.8,0.6,0)。上述几何特征中的位置即为该几何特征所描述对象的位置，可以通过坐标来表示。
本申请实施例中，可以先识别对象，再将对象的点云数据进行直线/曲线/平面/曲面拟合，得到确定对象的几何特征。应理解，本申请实施例中可以不需要获取环境中所有的物体的几何特征，而是获取环境中部分物体的几何特征，比如针对环境中固定不动的物体或物体上的部分进行几何特征的提取。本申请实施例中，需要获取几何特征的物体或物体上 的部分(本申请实施例中也称对象)可以是路沿、路面、交通标线(traffic index line)、交通杆、交通标志牌、车道线、树干、隧道、建筑物等。应理解，本申请实施例中，可以将一个物体称为一个对象，也可以将一个物体中的部分称为一个对象。一个对象对应一个几何特征。
例如，交通杆可以包括横杆和纵杆，其中，横杆可以对应一个直线的几何特征，纵杆可以对应一个直线的几何特征。
又例如，长方体状的建筑物可以包括上边缘、下边缘、左边缘、右边缘、主平面和一个侧面等，其中，上边缘、下边缘、左边缘、右边缘可以分别对应一个直线的几何特征，主平面和一个侧面可以分别对应一个平面几何特征。
又例如，树干可以对应一个直线的几何特征。
应理解，从点云数据中除可以提取出几何特征外，还可以识别点云数据中的对象的属性，例如，对象所属的分类、尺寸等。该属性有助于更精确进行几何特征的匹配，关于几何特征的匹配，将在其他部分中介绍，此处不再赘述。
其中，对象的类别可以以对象所属物体的类别来划分，包括路沿、路面、道路指示线、交通杆、交通标志牌、车道线、树干、建筑物等。在本申请的一种实现中，还可以对上述类别进一步地划分。例如，建筑物包括建筑物的平面、左边缘、右边缘、上边缘、下边缘等；又例如，交通杆可以包括横杆和纵杆等；又例如道路指示线可以包括实线、虚线、转向线等。在本申请实施例的另一种实现中，对象的类别也可以以对象的几何形状进行划分，此时对象的类别可以包括直线形状、曲线形状、平面形状和曲面形状，应理解，对象的类别还可以包括上述两种的组合或以其他方式进行分类等，此处不作限定。
其中，尺寸可以是线的长度、平面的面积等，本申请实施例不作限定。尺寸是针对该几何特征所描述对象的类别，可以对应于对象的长、宽、高或表面积等。
其中，位置可以是几何特征所描述对象的位置，由于位置与对象的一一对应，该位置也称为几何特征的位置。几何特征的位置可以是在点云数据所采用的坐标系下描述的对象的位置坐标，可以约定几何特征的位置坐标的选取规则。几何特征的位置坐标可以是该几何特征所对应对象的几何中心的位置坐标；或者，几何特征的位置坐标是该几何特征所对应对象最低处的位置坐标(即该对象的点云数据中z轴坐标最小的点对应的位置坐标)；或者，对象的位置坐标是该对象的点云数据中x轴和/或y轴坐标最小的点对应的坐标。需要说明的是，本申请实施例中也可以采用其他标准来设定几何特征所描述对象的位置，对此不作限定。
在本申请实施例的一种实现中，针对不同的分类的对象也可以采用不同的选取规则来确定该对象对应的几何特征的位置坐标。例如，针对建筑物的主平面，其几何特征为平面，该几何特征的位置坐标可以是该主平面的几何中心的位置坐标；又例如，针对树干，其集合特征为直线，几何特征的位置坐标可以是树干的点云数据中z轴坐标最小的点对应的位置坐标；又例如，针对建筑物的上边缘，其几何特征为直线，该几何特征的位置坐标可以是建筑物的上边缘的点云数据中最小x轴坐标的点对应的位置坐标。
需要指出的是，对于部分的对象如，路面、路沿等的几何特征的选取可以包括但不限于一下两种方式：
第一方式：选取对象的点云数据中的任意位置的坐标作为该对象对应的几何特征的位置。应理解，此时位置可以用于体现各个对象之间的相对位置关系。
第二方式：可以以其他对象作为参照，根据被参照对象的位置确定该对象对应的几何特征的位置。例如，对于路面的几何特征来说，其位置可以是设于道路两侧的两个对象的位置的中点。
其中，下面举例说明几何特征的表达方式：
如图1A所示的道路场景示意性说明图和图1B所示的几何特征的示意性说明图，图1A中a图、b图、c图、d图、e图为五个道路场景示意图，图1B中f图、g图、h图、i图、j图为分别从a图、b图、c图、d图、e图所示的道路场景的点云数据中提取到的几何特征的示意性说明图。
例如，由a图和f图所示，建筑楼的左边缘、右边缘、上边缘可以分别对应一个直线的几何特征，可以包括直线的方向向量和位置；建筑楼的墙面可以对应一个平面的几何特征，例如，包括法向量和平面几何中心的位置。
又例如，如b图和g图所示，交通杆中的横杆和纵杆可以分别对应的一个直线的几何特征。
又例如，如c图和h图所示，直线形状的路沿可以对应一个直线的几何特征。
又例如，如d图和i图所示，隧道内壁可以对应一个曲面的几何特征。
又例如，如e图和j图所示，支撑柱的多个墙体可以分别对应一个平面的几何特征。支撑柱的边缘可以分别对应一个直线的几何特征。
(6)几何特征地图
我们可以提取点云地图中的点云数据中的几何特征，点云地图中提取得到的所有的几何特征构成几何特征地图。也就是说，几何特征地图为从点云地图的点云数据中提取的几何特征所形成的地图。可选地，该几何特征地图还可以包括几何特征的属性，如尺寸、所属类别等，还可以包括位置和该位置对应的地址、道路的几何特征对应的道路名称等。该几何特征地图可以作为地图(如Google地图、百度地图、高德地图或点云地图等地图)中的一个图层，也可以单独作为地图。在车辆进行定位的过程中，车辆或车辆绑定的终端可以加载几何特征地图，也可以仅加载几何特征地图的数据，本申请实施例不作限定。
(7)几何特征的匹配
几何特征的匹配是指将不同坐标系下的两组几何特征进行匹配，进而建立不同坐标系中几何特征之间对应关系。在本申请实施例中，车辆实时采集的点云数据是基于车辆的预测位姿(即不准确的位置和姿态)确定的坐标系下的点的数据，几何特征的匹配的过程也就是将从该点云数据中提取到的几何特征与几何特征地图中的几何特征相匹配。应理解，建立对应关系的两个几何特征，其本质上是一个对象在不同坐标系下的表达。在对车辆进行定位之前，首先需要找到几何特征地图中与车辆采集的点云数据提取到的一组几何特征相匹配的一组几何特征。具体地，可以根据该两组几何特征的位置、方向和属性等中的至少一种来实现建立两组几何特征之间的匹配，具体实现可以参见下述定位方法中相关描述，此处不再赘述。
(8)通过几何特征进行定位的原理
在道路场景中，路沿、交通杆、交通标志牌、车道线、树干、建筑物等物体的3D几何特征在三维空间中的分布具有非平行、非共面特征。理论上，两个以上非共面的几何特征即可确定出两个坐标系之间的变换关系，得到车辆的定位结果。应理解，在定位过程中，越多的几何特征会得到越精确的定位结果。
(9)卡尔曼滤波(Kalman filter)
卡尔曼滤波是一种状态最优估计方法，假设一个离散线性动态系统，可以根据上一个时刻的状态推测下一个时刻的状态。预测方程可以表示为
Figure PCTCN2019105810-appb-000004
其中,下标“k”表示当前时刻，下标“k-1”表示上一时刻，状态中符号“^”表示该状态为估计值，状态中上标“ˉ”表示该状态是根据上一个状态的预测结果，
Figure PCTCN2019105810-appb-000005
表示当前时刻下状态的预测估计值，a k为当前时刻下车辆的控制输入，如加速度、转向等。所有的预测都是包含噪声的，噪声越大，不确定性越大，通过协方差矩阵(covariance)来表示状态预测噪声，通常用P来表示。噪声协方差矩阵的预测可表示为：P k ˉ＝A kP k-1A T+Q k，其中，A k状态转移矩阵，Q k表示系统模型本身的噪声矩阵。系统的观测值为z k，根据系统状态的预测结果
Figure PCTCN2019105810-appb-000006
可以预测出系统的观测值
Figure PCTCN2019105810-appb-000007
其中，
Figure PCTCN2019105810-appb-000008
H k为系统的观测矩阵。
状态最优估计值
Figure PCTCN2019105810-appb-000009
是通过观测值与观测的预测估计值之间的残差
Figure PCTCN2019105810-appb-000010
乘以卡尔曼系数(也称卡尔曼增益)K k更新得到，即
Figure PCTCN2019105810-appb-000011
其中，H k为观测矩阵
Figure PCTCN2019105810-appb-000012
表示实际观测值和预期观测值之间的残差。用该残差乘以系数K k可以对预测状态
Figure PCTCN2019105810-appb-000013
进行修正。其中，卡尔曼增益矩阵K k实际上表征了状态最优估计过程中系统状态和预测状态之间的方差P最小时的估计系数。更新最优状态估计的噪声协方差矩阵P，用于下一轮迭代时计算卡尔曼滤波增益，状态的不确定性减小；但在下轮迭代中由于传递噪声的引入，不确定性会增大，卡尔曼滤波就是在不确定性的变化中寻求一种最优的状态估计。
(10)精确度
在本申请实施例中，第一位姿、第二位姿、估计位姿、预测位姿等都是对车辆的位姿的测量值，各个位姿的精确度不同。本申请实施例中精确度是指车辆的测量位姿，如第一位姿、第二位姿、估计位姿、预测位姿等与车辆真实位姿的接近程度，即第一位姿的精确度指示第一位姿态和车辆的真实位姿的差距，第二位姿的精确度指第二位姿态和车辆的真实位姿的差距，精确度低指示差距大，精确度高指示差距小。本申请实施例的目的在于，通过对精确度低的第一位姿进行调整得到精确度高的第二位姿。也就是说，相对于第一位姿，第二位姿更接近于车辆的真实位姿，可以作为车辆的真实位姿(也称为实际位姿)，因此，可以将第一位姿称为预测位姿，第二位姿称为实际位姿。应理解，第二位姿也是对车辆的实际位姿的测量值，只是相对于第一位姿来说，更加接近车辆的实际位姿。
下面介绍本申请实施例提供的系统架构。
参见图2，本申请实施例提供了一种系统架构，该系统10所示，该系统10可以包括： 数据采集设备110、数据库120、点云地图服务器130、几何特征地图生成设备140、地图服务器150、车辆160、终端180和定位服务器190，其中，针对不同的应用场景，车辆160、终端180和定位服务器190均可以作为定位设备，本发明对于定位设备具体是车辆、终端或定位服务器不做限定。
数据采集设备110用于通过点云采集装置采集点云数据，并将点云数据存入数据库120，可以是配置有点云采集装置的车辆，或者其他可以实现点云数据采集的装置。在另一种实现中，数据采集设备110也可以提取采集到的点云数据中的几何特征，将几何特征存入数据块120。
点云地图服务器130用于将数据采集设备110采集到的点云数据形成点云地图，还可以接收来自定位设备的针对特定区域的点云数据的请求，响应该请求，以向定位设备发送该特定区域的点云数据。
几何特征地图生成设备140可以是服务器或者计算机等具有计算功能的设备，用于从数据库120或者点云地图服务器130中获取点云地图的点云数据，进而提取该点云数据中的几何特征，得到几何特征地图。进一步地，可以将几何特征地图存入数据库130，也可以存在本地。几何特征地图生成设备140还可以接收定位设备发送的针对第一区域的几何特征的获取请求，进而响应该请求，向定位设备发送第一区域的几何特征。应理解，几何特征地图的生成和响应定位设备的功能也可以分别通过不同的设备来实现，对此，本申请实施例不作限定。
地图服务器150可以是地图应用的服务器，例如百度地图的服务器、谷歌地图的服务器、高德地图的服务器等地图应用的服务器。定位设备可以与地图服务器150建立通信连接，进行数据交互，以实现定位设备可以根据当前的定位到的位姿导航至目的地。应理解，几何特征地图生成设备140也可以包括地址、道路等数据，定位设备也可以与几何特征地图生成设备140建立通信连接，以实现定位设备的导航。
定位设备可以是车辆160、车辆160上的计算机系统或与车辆160通行连接的终端180，如手机、平板电脑等；还可以是定位服务器190。车辆160上配置了点云采集装置、惯性传感器等。当定位设备为定位服务器190或终端180时，车辆190可以向定位服务器190或终端180发送采集到的当前环境的第一点云数据和第一位姿，定位服务器190/终端180根据接收到的第一点云数据和第一位置，对车辆160的第一位姿进行调整，并将定位得到的第一位姿发送至车辆160。
定位服务器190可以是为车辆提供定位需求的服务器。在一种实现场景中，定位服务器190可以获取车辆160通过点云采集装置采集的第一点云数据，也可以预存几何特征地图，实现对车辆的定位；进一步地，定位服务器还可以将定位得到的第二位姿发送至车辆160或终端180。
其中，上述点云采集装置具体可以是至少一个激光雷达，激光雷达可以是多线雷达激光传感器，例如4线、8线、16线、32线、64线或其他数量线束的激光雷达。点云采集装置也可以是立体摄像头、TOF摄像头等，其中，立体摄像头可以是包括多个摄像头，也可以是包括一个或多个摄像头和激光雷达的组合。惯性传感器可以包括但不限于陀螺仪、加速度计、磁力传感器等中一种或多种。
车辆的计算机系统可以将通过惯性传感器采集到的运动信息和通过点云采集装置采集到的点云数据、或者从该点云数据中提取得到的几何特征发送给定位设备，以使定位设备根据接收到的数据进行定位。
定位设备可以向几何特征地图生成设备140请求获取几何特征地图，也可以从几何特征地图生成设备140下载几何特征地图，进而将通过点云采集装置采集到的点云数据中提取的几何特征与几何特征地图中的几何特征进行匹配，进一步地，利用匹配得到的几何特征进行定位。具体地定位方法是，定位设备可以确定车辆的第一位姿，该第一位姿是对车辆当前位置的估算值，精确度较低；定位设备获取车辆上的点云采集装置采集到的点云数据，该点云数据是基于第一位姿确定的坐标系下的描述；进一步地，从该点云数据中提取得到多个第一几何特征，再从几何特征地图中查找与该多个第一几何特征相匹配的多个第二几何特征，其中，第一几何特征与第二几何特征是一一对应的，进而，由于多个第一几何特征是基于不准确的第一位姿对该多个第二几何特征所描述对象的表达，故，定位设备可以根据多个第一几何特征和多个第二几何特征，得到车辆的准确的位姿，即第二位姿。
应理解，上述点云地图服务器130、地图服务器150不是本申请实施例中系统10必须的组成部分，系统10还可以包括其他设备，此处不作限定。
下面介绍本申请实施例提供的车辆。
图3是本发明实施例提供的车辆的功能框图。在一个实施例中，将车辆配置为完全或部分地自动驾驶模式。例如，车辆100可以在处于自动驾驶模式中的同时控制自身，并且可通过人为操作来确定车辆及其周边环境的当前状态，确定周边环境中的至少一个其他车辆的可能行为，并确定该其他车辆执行可能行为的可能性相对应的置信水平，基于所确定的信息来控制车辆100。在车辆100处于自动驾驶模式中时，可以将车辆100置为在没有和人交互的情况下操作。
车辆100可以图2中所示的系统中车辆160，可包括各种子系统，例如行进系统102、传感系统104、控制系统106、一个或多个外围设备108以及电源111、计算机系统112和用户接口116。可选地，车辆100可包括更多或更少的子系统，并且每个子系统可包括多个元件。另外，车辆100的每个子系统和元件可以通过有线或者无线互连。
行进系统102可包括为车辆100提供动力运动的组件。在一个实施例中，推进系统102可包括引擎118、能量源119、传动装置122和车轮/轮胎121。引擎118可以是内燃引擎、电动机、空气压缩引擎或其他类型的引擎组合，例如气油发动机和电动机组成的混动引擎，内燃引擎和空气压缩引擎组成的混动引擎。引擎118将能量源119转换成机械能量。
能量源119的示例包括汽油、柴油、其他基于石油的燃料、丙烷、其他基于压缩气体的燃料、乙醇、太阳能电池板、电池和其他电力来源。能量源119也可以为车辆100的其他系统提供能量。
传动装置122可以将来自引擎118的机械动力传送到车轮121。传动装置122可包括变速箱、差速器和驱动轴。在一个实施例中，传动装置122还可以包括其他器件，比如离合器。其中，驱动轴可包括可耦合到一个或多个车轮121的一个或多个轴。
传感系统104可包括感测关于车辆100周边的环境的信息的若干个传感器。例如，传 感系统104可包括全球定位系统122(全球定位系统122可以是GPS系统，也可以是北斗系统或者其他定位系统)、惯性测量单元(inertial measurement unit，IMU)124、点云采集装置126，其中，点云采集装置126可以包括激光雷达127、立体摄像头128、TOF摄像头等。传感系统104还可包括被监视车辆100的内部系统的传感器(例如，车内空气质量监测器、燃油量表、机油温度表等)。来自这些传感器中的一个或多个的传感器数据可用于检测对象及其相应特性(位置、形状、方向、速度等)。这种检测和识别是自主车辆100的安全操作的关键功能。
点云采集装置126可用于获取车辆100周围环境的点云数据，以实现对车辆100的地理位置和姿态的精确估计。车辆100中的全球定位系统122，例如GPS、BPS等，可以用于对车辆100的地理位置进行粗略估计。其中，IMU 124用于基于惯性加速度来感测车辆100的位置和朝向变化，可以用于粗略估计车辆的姿态。在一个实施例中，IMU 124可以是加速度计和陀螺仪的组合。
点云采集装置126具体可以是至少一个激光雷达127，激光雷达127可以是多线激光雷达，例如4线、8线、16线、32线、64线或其他数量线束的激光雷达。激光雷达127可利用无线电信号来感测车辆100的周边环境内的物体。在一些实施例中，除了感测物体以外，点云采集装置126还可用于感测物体的速度和/或前进方向。点云采集装置126也可以是立体摄像头128，其中，立体摄像头128可以是包括多个摄像头、也可以是包括一个或多个摄像头和激光雷达传感器。立体摄像头可用于捕捉车辆100的周边环境的多个包括深度信息的图像。
控制系统106为控制车辆100及其组件的操作。控制系统106可包括各种元件，其中包括转向系统132、油门134、制动单元136、传感器融合算法138、计算机视觉系统141、路线控制系统142以及障碍物避免系统144。
转向系统132可操作来调整车辆100的前进方向。例如在一个实施例中可以为方向盘系统。
油门134用于控制引擎118的操作速度并进而控制车辆100的速度。
制动单元136用于控制车辆100减速。制动单元136可使用摩擦力来减慢车轮121。在其他实施例中，制动单元136可将车轮121的动能转换为电流。制动单元136也可采取其他形式来减慢车轮121转速从而控制车辆100的速度。
计算机视觉系统141可以操作来处理和分析由相机130捕捉的图像以便识别车辆100周边环境中的物体和/或特征。所述物体和/或特征可包括交通信号、道路边界和障碍物。计算机视觉系统141可使用物体识别算法、运动中恢复结构(Structure from Motion，SFM)算法、视频跟踪和其他计算机视觉技术。在一些实施例中，计算机视觉系统141可以用于为环境绘制地图、跟踪物体、估计物体的速度等等。
路线控制系统142用于确定车辆100的行驶路线。在一些实施例中，路线控制系统142可结合来自传感器138、GPS 122和一个或多个预定地图的数据以为车辆100确定行驶路线。
障碍物避免系统144用于识别、评估和避免或者以其他方式越过车辆100的环境中的潜在障碍物。
当然，在一个实例中，控制系统106可以增加或替换地包括除了所示出和描述的那些以外的组件。或者也可以减少一部分上述示出的组件。
车辆100通过外围设备108与外部传感器、其他车辆、其他计算机系统或用户之间进行交互。外围设备108可包括无线通信装置146、车载电脑148、麦克风151和/或扬声器152。
在一些实施例中，外围设备108提供车辆100的用户与用户接口116交互的手段。例如，车载电脑148可向车辆100的用户提供信息。用户接口116还可操作车载电脑148来接收用户的输入。车载电脑148可以通过触摸屏进行操作。在其他情况中，外围设备108可提供用于车辆100与位于车内的其它设备通信的手段。例如，麦克风151可从车辆100的用户接收音频(例如，语音命令或其他音频输入)。类似地，扬声器152可向车辆100的用户输出音频。
无线通信装置146可以直接地或者经由通信网络来与一个或多个设备无线通信。例如，无线通信装置146可使用3G蜂窝通信，例如CDMA、EVD0、GSM/GPRS，或者4G蜂窝通信，例如LTE。或者5G蜂窝通信。无线通信装置146可利用WiFi与无线局域网(wireless local area network，WLAN)通信。在一些实施例中，无线通信装置146可利用红外链路、蓝牙或ZigBee与设备直接通信。无线通信装置146也可包括一个或多个专用短程通信(dedicated short range communications，DSRC)设备，这些设备可实现车辆和/或路边单元(road side unit，RSU)之间的公共和/或私有数据通信。
电源111可向车辆100的各种组件提供电力。在一个实施例中，电源111可以为可再充电锂离子或铅酸电池。这种电池的一个或多个电池组可被配置为电源为车辆100的各种组件提供电力。在一些实施例中，电源111和能量源119可一起实现，例如一些全电动车中那样。
车辆100的部分或所有功能受计算机系统112控制。计算机系统112可包括至少一个处理器113，处理器113执行存储在例如数据存储装置114这样的非暂态计算机可读介质中的指令115。计算机系统112还可以是采用分布式方式控制车辆100的个体组件或子系统的多个计算设备。
处理器113可以是任何常规的处理器，诸如商业可获得的CPU。替选地，该处理器可以是诸如ASIC或其它基于硬件的处理器的专用设备。尽管图3功能性地示了处理器、存储器、和在相同块中的计算机110的其它元件，但是本领域的普通技术人员应该理解该处理器、计算机、或存储器实际上可以包括可以或者可以不存储在相同的物理外壳内的多个处理器、计算机、或存储器。例如，存储器可以是硬盘驱动器或位于不同于计算机110的外壳内的其它存储介质。因此，对处理器或计算机的引用将被理解为包括对可以或者可以不并行操作的处理器或计算机或存储器的集合的引用。不同于使用单一的处理器来执行此处所描述的步骤，诸如转向组件和减速组件的一些组件每个都可以具有其自己的处理器，所述处理器只执行与特定于组件的功能相关的计算。
在此处所描述的各个方面中，处理器可以位于远离该车辆并且与该车辆进行无线通信。在其它方面中，此处所描述的过程中的一些在布置于车辆内的处理器上执行而其它则由远程处理器执行，包括采取执行单一操纵的必要步骤。
在一些实施例中，数据存储装置114可包含指令115(例如，程序逻辑)，指令115可被处理器113执行来执行车辆100的各种功能，包括以上描述的那些功能。数据存储装置114也可包含额外的指令，包括向推进系统102、传感系统104、控制系统106和外围设备108中的一个或多个发送数据、从其接收数据、与其交互和/或对其进行控制的指令。
除了指令115以外，数据存储装置114还可存储数据，例如通过点云采集装置126采集的第一点云数据、第一点云数据提取出的第一几何特征、几何特征地图、道路地图、路线信息，车辆的位置、方向、速度以及其它这样的车辆数据，以及其他信息。这种信息可在车辆100在自主、半自主和/或手动模式中操作期间被车辆100和计算机系统112使用。
本申请实施例中，车辆100或计算机系统112可以基于定位系统122例如全球定位系统122和惯性测量单元124确定车辆的第一位姿，以粗略估算车辆当前位置；进而获取车辆上的点云采集装置126采集到的第一点云数据；进一步地，从该第一点云数据中提取得到多个第一几何特征，该第一几何特征是在基于第一位姿确定的坐标系下对多个对象的描述，再从几何特征地图中查找与该多个第一几何特征相匹配的多个第二几何特征，其中，第一几何特征与第二几何特征是一一对应的，进而，由于多个第一几何特征是基于不准确的第一位姿对该多个第二几何特征所描述对象的表达，故而，车辆100或计算机系统112可以根据多个第一几何特征和多个第二几何特征，得到车辆的准确的位姿，即第二位姿。
用户接口116，用于向车辆100的用户提供信息或从其接收信息。可选地，用户接口116可包括在外围设备108的集合内的一个或多个输入/输出设备，例如无线通信装置146、车车在电脑148、麦克风151和扬声器152。
计算机系统112可基于从各种子系统(例如，行进系统102、传感系统104和控制系统106)以及从用户接口116接收的输入来控制车辆100的功能。例如，计算机系统112可利用来自控制系统106的输入以便控制转向单元132来避免由传感系统104和障碍物避免系统144检测到的障碍物。在一些实施例中，计算机系统112可操作来对车辆100及其子系统的许多方面提供控制。
可选地，上述这些组件中的一个或多个可与车辆100分开安装或关联。例如，数据存储装置114可以部分或完全地与车辆1100分开存在。上述组件可以按有线和/或无线方式来通信地耦合在一起。
可选地，上述组件只是一个示例，实际应用中，上述各个模块中的组件有可能根据实际需要增添或者删除，图3不应理解为对本发明实施例的限制。
在道路行进的自动驾驶汽车，如上面的车辆100，可以识别其周围环境内的物体以确定对当前速度的调整。所述物体可以是其它车辆、交通控制设备、或者其它类型的物体。在一些示例中，可以独立地考虑每个识别的物体，并且基于物体的各自的特性，诸如它的当前速度、加速度、与车辆的间距等，可以用来确定自动驾驶汽车所要调整的速度。
可选地，自动驾驶汽车车辆100或者与自动驾驶车辆100相关联的计算设备(如图3的计算机系统112、计算机视觉系统141、数据存储装置114)可以基于所识别的物体的特性和周围环境的状态(例如，交通、雨、道路上的冰、等等)来预测所述识别的物体的行为。可选地，每一个所识别的物体都依赖于彼此的行为，因此还可以将所识别的所有物体全部一起考虑来预测单个识别的物体的行为。车辆100能够基于预测的所述识别的物体的行为 来调整它的速度。换句话说，自动驾驶汽车能够基于所预测的物体的行为来确定车辆将需要调整到(例如，加速、减速、或者停止)什么稳定状态。在这个过程中，也可以考虑其它因素来确定车辆100的速度，诸如，车辆100在行驶的道路中的横向位置、道路的曲率、静态和动态物体的接近度等等。
除了提供调整自动驾驶汽车的速度的指令之外，计算设备还可以提供修改车辆100的转向角的指令，以使得自动驾驶汽车遵循给定的轨迹和/或维持与自动驾驶汽车附近的物体(例如，道路上的相邻车道中的轿车)的安全横向和纵向距离。
上述车辆100可以为轿车、卡车、摩托车、公共汽车、船、飞机、直升飞机、割草机、娱乐车、游乐场车辆、施工设备、电车、高尔夫球车、火车、和手推车等，本发明实施例不做特别的限定。
下面结合附图详细描述本申请实施例。
实施例(一)：
首先介绍本申请实施例涉及的几何特征的提取方法，该方法可以由图2中车辆160、定位设备、几何特征地图生成设备140或图3中车辆100执行，本申请实施例以执行主体为车辆为例来说明。请参阅图4A所示的几何特征提取方法，该方法可以包括但不限于如下步骤：
S402：获取待处理的点云数据。
在车辆进行定位的应用场景中(本申请实施例三)，该待处理的点云数据为车辆通过点云采集装置获取的周围环境的第一点云数据。从该第一点云数据中提取出的几何特征，即为实施例三中的多个第一几何特征或N个第一几何特征，提取得到的几何特征用于车辆的定位。此时，该第一点云数据可以是一帧点云数据也可以是多帧点云数据，还可以是对多帧点云数据进行帧间点云叠加后得到的点云数据，对此，本申请实施例不作限定。
应理解，对于定位设备来说，若定位设备为终端或服务器，则定位设备可以与车辆建立通信连接，进而接收车辆通过点云采集装置获取的第一点云数据。
在生成几何特征地图的应用场景中(本申请实施例二)，该待处理的点云数据为几何特征地图生成设备从数据库、数据采集设备或点云地图中获取的第二点云数据，从该待处理的点云数据中提取的几何特征用于生成几何特征地图。此时，几何特征地图生成设备获取的待处理的点云数据可以是待处理区域的点云数据，也可以是待处理的一帧或多帧点云数据，还可以是多帧点云数据进行帧间点云叠加后得到的点云数据，对此本申请实施例不作限定。
可选地，在S404之前，车辆还可以对待处理的点云数据进行降噪处理，例如，滤除点云数据中的离群值，以滤除噪声过大的干扰点，只保留有效的数据点，提高几何特征提取的准确度；又例如，车辆可以对点云数据进行降采样，即减少点云数据中采样点个数，以减少数据量，降低数据处理量，提高车辆的定位速度。
S404：提取待处理的点云数据中的几何特征。
在S404的一种实现中，S404可以包括如下步骤：
S4042：识别待处理的点云数据中的至少一个对象。
本申请实施例中识别待处理的点云数据中的多个对象包括但不限于识别待处理的点云数据中对象的属性、该属性可以是对象所述的类别。在本申请实施例的一种实现中，该对象所属的类别(第一类别)包括直线、曲线、平面、曲面等。在本申请实施例的另一种实现中，该对象所属的类别(第二类别)包括路沿、路面、树干，建筑物的平面、左边缘、右边缘、上边缘、下边缘，交通杆的横杆、纵杆，道路指示线的实线、虚线、转向线等。
在本申请实施例的一种实现中，对象可以不包括物理含义，车辆仅识别出其几何形状，进而采用该几何形状对应的技术方法提取其几何特征。在本申请实施例的另一种实现中，对象可以包括具体的物理含义，比如路沿、路面、道路指示线、交通指示牌、道路指示线、树干、建筑物的平面、左边缘、右边缘、上边缘、下边缘，交通杆的横杆、纵杆，道路指示线的实线、虚线、转向线等，本申请实施例不作限定。
S4044：根据第一对象的点云数据，确定第一对象的几何特征，第一对象为识别到的多个对象中的任意一个对象。
本申请实施例中，首选，识别待处理的点云数据中的对象的几何形状，进一步地，可以从待处理的点云数据中划分出各个对象的点云数据，进而，对各个对象的点云数据进行直线/曲线/平面/曲面等拟合，如直线/曲线/平面/曲面，对象的几何表达的几何特征即为该对象的几何特征。
若第一对象的几何形状为直线，对第一对象的点云数据进行直线拟合，得到第一对象的第一几何特征，第一对象的第一几何特征为拟合得到的直线的几何特征；若第一对象的几何形状为曲线，对第一对象的点云数据进行曲线拟合，得到第一对象的第一几何特征，第一对象的第一几何特征为拟合得到的曲线的几何特征；若第一对象的几何形状为平面，对第一对象的点云数据进行平面拟合，得到第一对象的第一几何特征，第一对象的第一几何特征为拟合得到的平面的几何特征；若第一对象的几何形状为曲面，对第一对象的点云数据进行曲面拟合，得到第一对象的第一几何特征，第一对象的第一几何特征为曲面拟合得到的曲线的几何特征。
应理解，对于第二类别，车辆可以根据类别与几何形状的对应关系，确定各个类别所采用的那种拟合方式。例如，针对线状对象，如类别为建筑物的左边缘、右边缘、上边缘、下边缘，交通杆的横杆、纵杆，指示牌的杆、树干等，采用直线拟合的方式计算其几何特征。又例如，针对平面状的对象，如建筑物的平面、广告牌的牌面等采用平面拟合的方式计算其几何特征。
下面举例介绍各个对象的提取原理和方法：
(1)路面和路沿
请参阅图4B和图4C所示提取路面和路沿的几何特征的原理示意图，如图4B所示的路沿的点云数据的分布，路面通常为平坦的平面，路沿可以包括上边沿和下边缘，其中，在上边缘的点云数据和下边缘的点云数据在高度上都具有突变。对于多线激光雷达来说，在车辆的运动方向上，激光雷达发出的激光束，如图4B中激光束i或图4C中激光束j，横跨路面和路沿，因而，每一个激光束扫描得到的点云数据都可以得到观测到高度的变化。因此，车辆可以根据每一个激光束得到的点云数据的在高度的变化，确定路面和路沿的边界(下边缘)和路沿的上边缘。例如，在上边缘和下边缘的高度差的Δh大于第一高度的 情况下，确定为该上边缘和下边缘之间的区域为路沿，进一步地，基于上边缘和下边缘的点云数据，得到路沿的几何特征。
在本申请实施例的一种实现中，路面的几何特征可以是路面所在平面的法向量，如图4B和图4C中V pavement。
在本申请实施例的又一种实现中，可以将路沿的上边缘/下边缘所在直线或曲线的几何特征作为路沿的几何特征，如图4B中中向量V curb_up或向量V curb_low即为路沿的几何特征。如图4C中，路面和路沿的边界所在曲线(即图中曲线f curb_low)上的多个坐标点的组合或多个坐标点上的法向量的组合作为曲线的几何特征，即(a点坐标，a点的法向量v a，b点坐标，b点法向量v b，c点坐标，c点法向量v c)。
在本申请实施例的又一种实现中，也可以将路沿的上边缘所在的直线/曲线的几何特征和下边缘所在的直线/曲线的几何特征的平均值作为路沿的几何特征，即，路沿的几何特征为(V curb_low+V curb_low)/2，本申请实施例不作限定。
在本申请实施例中，可以路沿的几何特征还可以包括路沿的位置，其中，路沿的位置可以是路沿的点云数据中的任意位置的坐标作为该对象对应的几何特征的位置。应理解，还可以包括其他规则来确定路沿的位置，此处不作限定。
本申请实施例中，车辆还可以标注该几何特征的属性、如尺寸、所属分类等。
可见，本申请实施例中，通过向量、位置和属性来替代对象上大量的点云数据，与原始的点云数据相比，该几何特征占据更小的存储空间，大大降低数据的存储量，进一步地降低数据计算复杂度，以满足车辆定位对实时性的要求。
(2)道路指示线
由于道路指示线的涂层材料与路面的材料不同，道路指示线对激光的反射强度远远大于路面。故而，车辆可以提取路面的点云数据中对激光的反射将强度大于预设强度的点，得到道路指示线的点云数据，进一步地，根据道路指示线的点云数据得到道路指示线的几何特征。
请参阅图4D提取道路指示线的几何特征的几何特征的原理示意图，道路指示线的几何特征可以是道路指示线所在直线或曲线的几何特征。各个道路指示线的几何特征可以如4D中所示的V1-V10。
在本申请实施例的另一种实现中，若点云数据是通过立体摄像头获取的，则可以基于图像识别技术来识别点云数据中的道路标志线，进一步地，根据道路指示线的点云数据得到道路指示线的几何特征。在一些实施例中，若道路指示线破损，被污染等，造成车道线丢失，车辆可以利用图像增强技术，还原道路指示线。在一些实施例中，由于交通拥堵导致激光雷达发送的线束被其他车辆遮挡，导致可以测量得到的车道指示线的信息缺失，此时，车辆可以根据上一时刻的得到的车道指示线的几何特征和车辆的运动轨迹来估计当前时刻的车道指示线的几何特征。
在一些实施例中，左转、右转、直行、掉头等道路指示线可以作为几何特征的属性，辅助几何特征的匹配。
(3)交通杆、树干、路牌或广告牌等杆状对象
在本申请实施例的一种实现中，车辆可以以三维网格为标尺对点云数据进行划分，该 三维网格的尺寸(X轴长度×Y轴长度×Z轴长度)可以是0.2m×0.2m×0.2m，或其他数值，对此不作限定。应理解，当三维网格内存在点云数据时，则该三维网格内存在物体。对该三维网格内物体的高度进行编码，得到高度分布图。标记为(i，j，k)三维网格表示为X方向上的第i个、Y方向上处于的第j个以及Z方向上的第k个确定的三维网格，该三维网格(i，j，K)被编码的高度值H为：
Figure PCTCN2019105810-appb-000014
其中，在标识在三维网格(i，j，k)中存在物体时，N(k)取值为1，否则，取值为0，其中，i、j、k为正整数。
如图4E所示的树木的点云数据的分布示意图，三维网格(4，3，5)被编码的高度值大于第一阈值，图4E中K为5，三维网格(4，3，5)上层(即Z＝4.0以上)的三维网络中，在位置(4，3)周围，三维网格的高度值迅速向周围三维网格扩张，如图4E中Z＝4.2所示的点云数据分布，则可以判断该物体为树木。三维网格(4，3，5)对应的高度值即为树干的高度。根据位于(4，3)的三维网格内的点云数据，进行直线拟合，计算出树干的方向向量、确定树干位置等。
应理解，路灯杆、广告牌、交通杆、树木等物体随着Z方向的增大，邻域内的被编码的高度值具有不同的特性，据此可以区分出路灯杆、广告牌等杆状物体。
如图4F所示的广告牌的几何特征的示意性说明图，根据不同高度下点云数据的分布，可以识别出该点云描述物体为广告牌。进一步地，针对广告牌的牌面，可根据多个高度z的位于(4，3)的三维网格内的点云数据的分布，提取广告牌的杆的方向向量；针对广告牌的杆，根据高度Z为2.2-3.2区间内的点云数据的分布图，提取广告牌内部的点云，将广告牌内部点云进行平面拟合，提取广告牌的牌面的法向量。
如图4G所示的路灯杆的几何特征的示意性说明图，根据不同高度点云数据的分布，可以识别出该点云描述物体为路灯杆，进一步地，根据位于(4，3)的三维网格内的点云数据，计算出路灯杆的方向向量、确定路灯杆的位置等。
在本申请另一实施例中，多线激光雷达的中的单束激光绕雷达旋转中心扫描时，测量距离以极坐标形式表示，如图4H所示指示牌上点的距离和角度的示意性说明图。当单束激光相邻两个角度返回的距离差大于距离阈值，认为该两个角度之间为物体的边界，结合相邻的激光线束得到的点可以分割出物体的测量点。进一步地，基于分割得到的测量点，进行直线、曲线、平面或曲面的拟合，得到物体上各个部分的几何特征。
上述列举的实施例仅为说明该方法的实现过程进行的示例性描述，在实际实施时，可以对上述各种实现进行不同是形式的组合、优化和各种方法的变换等以达到提取几何特征的目的。
(4)建筑物的外缘、墙面
如图4I所示为建筑物的点云数据的分布示意图，通过上述(3)中识别杆状物体的方式，车辆可以识别建筑物的外缘(左边缘、右边缘、上边缘、下边缘等)并分割出建筑物各个外缘的点云数据，进而，通过直线/曲线拟合等方式确定各个外缘的几何特征为该外缘所在的直线/曲线的几何特征。通过上述(3)中识别广告牌的牌面的方式，车辆可以识别出建筑的墙面，进而通过平面/曲面拟合等方式确定建筑物的墙面的几何特征为该建筑物的墙面所在的平面/曲面的几何特征。
应理解，其他对象的识别、对象的点云数据分割、几何特征的生成的原理可以参照上述各个对象中相关描述，本申请实施例不再例举。
实施例(二)
本申请实施例介绍一种地图生成方法，该方法应用上述实施例(一)所述的几何特征的提取方法，可以由图2中几何特征地图生成设备140执行，请参阅图5所示的几何地图的生成方法，该方法可以包括但不限于如下步骤：
S502：从点云地图中获取第二点云数据，该第二点云数据可以是点云地图中待处理区域的点云数据，也可以是点云地图中待处理的一帧或多帧点云数据，还可以是上述多帧点云数据进行帧间点云叠加后得到的点云数据，对此本申请实施例不作限定。应理解，几何特征地图生成设备可以以点云数据所在的区域对点云地图中的点云数据进行划分，
S504：提取该第二点云数据中的几何特征，点云地图中所有的第二点云数据提取得到的几何特征构成几何特征地图。
从该第二点云数据中提取的几何特征用于生成几何特征地图。此时，具体实现可以参见上述实施例(一)中从第一点云数据中提取几何特征的实现方式的相关描述，此处不再赘述。
可选地，该几何特征还可以包括属性，如所属类别、尺寸等信息。
几何特征地图中的几何特征可以编码成各种形式的存储格式，如图像、XML、文本或表格等，几何特征地图可以保存在几何特征地图生成设备或数据库中。
例如，几何特征地图可以以图像格式存储，图像中每一个像素点具有确定的位置，当该位置存在几何特征时，该像素点的值可以编码为该存在的几何特征的向量、尺寸或所属类别等。
可选地，该几何特征地图还可以包括位置和该位置对应的地址、道路的几何特征对应的道路名称等。该几何特征地图可以作为地图(如Google地图、百度地图、高德地图或点云地图等地图)中的一个图层，也可以单独作为地图，对此，不作限定。
可见，通过该方法得到的几何特征地图，通过向量、位置和属性来替代对象上大量的点云数据，与原始的点云地图相比，该几何特征地图占据更小的存储空间，大大降低数据的存储量，降低后续车辆定位过程数据计算复杂度，满足车辆定位对实时性的要求。
请参阅图6所示的定位原理示意图，其目的在于得到精确高的车辆的位置和姿态如图6所示，车辆采集到的第一点云数据提取得到的第一几何特征(虚线所示的几何特征)与几何特征地图中与该第一几何特征相匹配的第二几何特征(实线所示的几何特征)实质上是同一对象在不同坐标系下的表达。由于，车辆对当前位姿估计的第一位姿不准确导致其在根据第一位姿所确定的第一坐标系与几何特征地图中第二坐标系(即世界坐标系)的不同，即两个坐标系需要通过旋转和平移才能重合。理论上，若车辆估计的第一位姿非常精确，则第一坐标系和第二坐标系重合，采集得到的对象的第一几何特征与该对象在几何特征地图中的第二几何特征相同，即重合。因此，可以通过车辆采集得到的第一几何特征和第二几何特征之间的差异来调整车辆的第一位姿，得到车辆精确的位置，即第二位姿。
下面介绍本申请实施例涉及的定位方法，定位方法可以由定位设备执行，其中，该定位设备可以是车辆或车辆中的计算机系统；也可以是与车辆进行通信连接的终端，如手机、平板电脑等；也可以是定位芯片或定位装置，还可以是服务器或云端等。本申请实施例以执行设备为定位设备为例进行介绍。下述实施例(四)、实施例(五)和实施例(六)所示的定位方法，可以基于实施例(三)所示的定位方法来实现。
实施例(三)
请参阅图7所示的定位方法的流程示意图，该方法可以包括但不限于如下步骤：
S72：获取车辆通过点云采集装置采集的第一点云数据。
S74：从第一点云数据中提取N个第一几何特征，N为正整数。
S76：根据N个第一几何特征对车辆的第一位姿进行调整以得到车辆的第二位姿，第二位姿的精确度高于第一位姿态的精确度。
在本申请实施例的另一种实现中，定位设备也可以接收车辆或终端发送的上述N个第一几何特征，该N个第一几何特征是从车辆通过点云采集装置采集的第一点云数据中提取得到的。
上述方法提供了一种定位方法，通过从点云采集装置中采集到的第一点云数据中提取出的N个第一几何特征对精确度低的第一位姿进行纠正，可得到精确度高的第二位姿，相对于现有技术中的点云数据，本申请实施例采用数据量少的几何特征进行定位，大大减小了数据的运算量，使得车辆定位的耗时少，定位的实时性好。
本申请实施例的目的在于得到精确度高的车辆的位置和姿态，即本申请实施例中第二位姿。其中，第一位姿的精确度低于第二位姿的精确度，第一位姿为车辆的精确度低的车辆的位置和姿态，即包括第一位置和第一姿态。S72之前该方法还可以包括，S71，获取车辆的第一位姿，其中，车辆的第一位姿的获取方式可以包括但不限于如下三种实现方式：
实现方式A：
定位设备通过定位系统确定车辆的第一位置，该定位系统可以是全球定位系统(global positioning system，GPS)、北斗卫星导航系统(Beidou navigation satellite system，BDS)、基站定位(也称为移动位置服务(location based service，LBS))、或室内定位系统等。
第一姿态可以是车辆通过惯性传感器测量得到的车辆的俯仰角、偏航角和翻滚角，车辆可以基于该第一位姿确定第一坐标系，进而在该第一坐标系下描述通过点云采集装置采集到的第一点云数据。其中，惯性传感器可以包括加速度计、角速度传感器(如陀螺仪)和磁力传感器等。
实现方式B：
本实现方式中，定位设备可以通过本申请实施例中定位方法在上一时刻(k-1)定位到的第二位姿和车辆的控制输入等输入到车辆的动力学方程，预测车辆在当前时刻(k)的第一位姿。
实现方式C：
具体地，定位设备可以根据车辆在T1时刻的第二位姿和车辆上的惯性传感器在T1时刻到T2时刻检测到的运动轨迹确定车辆在T2时刻的第一位姿，其中，T2时刻为点云采集装置采集第一点云数据的时刻。
应理解，不限于上述3中实现方式，本申请实施例中获取第一位姿还可以包括其他实现方式，此处不作限定。
还应理解，在车辆获取到第一位姿之后，车辆可以通过点云采集装置采集当前环境的第一点云数据；进而，将采集到的第一点云数据发送至定位设备，使得定位设备在获取到第一点云数据后执行S74，得到N个第一几何特征。其中，该第一点云数据为在由第一位姿确定的空间中表示的车辆观测到的对象的表面上的点的信息，N个第一几何特征中每个第一几何特征用于指示在由第一位姿确定的空间(也称为第一坐标系描述的空间)中车辆观测到的一个对象的几何特征。在本申请实施例的另一种实现中，也可以由车辆执行步骤S74，进而，车辆将从第一点云数据中提取到的N个第一几何特征发送至定位设备。
还应理解，第一坐标系为车辆基于第一位姿确定的坐标系(空间)，由于第一位姿的精确度不高，该第一坐标系与世界坐标系之间存在偏差。
还应理解，第一点云数据的采集为现有技术，可参见现有技术中点云数据的采集、生成等方式中相关描述，此处不作限定。
其中，定位设备从第一点云数据中提取N个第一几何特征的具体实现可以参见实施例一中相关描述，本申请实施例不再赘述。本申请实施例，以从第一点云数据中提取了N个第一几何特征为例来说明，N为大于1的正整数，该N个第一几何特征可以表示为：V＝{V 1,V 2,…V N}。
在本申请实施例的一种实现中，在S76之前，该方法还可以包括：
S75：在几何特征地图中查找与N个第一几何特征相匹配的N个第二几何特征，第一几何特征与所述第二几何特征一一对应。应理解，几何特征地图可以包括在世界坐标系下描述的对象的几何特征。进而，定位设备可以基于车辆观测到的N个第一几何特征和几何特征地图中的N个第二几何特征对第一位姿进行调整以得到第二位姿。
几何特征地图包括地图上各个区域的几何特征。在一种实现中，定位设备缓存了几何特征地图，为提高匹配效率，定位设备可以从几何特征地图的第一区域中查找与N个第一几何特征相匹配的N个第二几何特征，如图6所示的第一几何特征和第二几何特征，其中，第一区域可以是基于第一位姿确定的区域，该第一区域可以大于车辆的点云采集装置的扫描范围。在另一种实现中，定位设备可以向几何特征地图生成设备或者几何特征地图生成设备请求获取几何特征地图中的第一区域中的几何特征，对此不作限定。
其中，确定第一区域的一种实现可以是：定位设备可以根据第一位置和该第一位置处的道路确定车辆在第一位置所处的场景，例如，道路交叉口、多层道路等的不同，通过不同的方式划定第一区域。其主要实现如下：
请参阅图8A所示的第一区域的示意性说明图，若第一位置位于道路上非交叉口的区域，第一区域可以如图8A所示的椭圆形区域801，第一位置802位于椭圆形区域801内。可以理解，第一区域还是以第一位置为中心、第一长度为半径确定的球形区域或圆形区域；也可以是以第一位置为中心，确定的长方体区域或矩形区域；还可以是点云采集装置在第一位置时，点云采集装置的探测范围。
请参阅图8B所示的第一区域的示意性说明图，若第一位置位于道路交叉口，比如第一道路与第二道路的交叉口的情况下，第一区域可以如图8B所示的区域803，第一位置804 位于区域803内。此时，由于第一位姿的不准确，第一区域需要考虑到各个道路上的对象，以防止划分出的几何特征地图中第一区域不能完全覆盖到N个第一几何特征所描述的对象。
请参阅图8C所示的第一区域的示意性说明图，若第一位置806处存在多层道路时，第一区域可以是如图8C所示的区域805。在一种实现中，定位设备可以先确定车辆所在的道路层，进而，确定第一区域805。其中，定位设备可以根据车辆导航要到达的目的地、或者车辆上一时刻定位得到的位姿和运动路径、第一位姿中的高度等来确定车辆所在的道路。进而，可以通过上述8A或者8B所示的方法确定第一区域。
在确定第一区域之后，可以将几何特征地图中第一区域内的几何特征与从采集到的第一点云数据中提取得到的N个第一几何特征进行匹配，以得到与N个第一几何特征相匹配的N个第二几何特征。其中，第一几何特征与第二几何特征一一匹配。本申请实施例以第一区域包括M个几何特征，定位设备从采集到的第一点云数据中提取得到的N个第一几何特征为例来说明，其中，M，N为正整数，M≥N，匹配过程的可以包括但不限于如下三种实现方式：
实现方式一：
定位设备可以将第一几何特征与M个几何特征逐个进行匹配，M个几何特征中与第一几何特征偏差最小的几何特征，即为与该第一几何特征相匹配的第二几何特征(本申请中也称该第一几何特征对应的第二几何特征)。
具体地，定位设备针对第一几何特征V i，计算第一几何特征V i与第一区域中每一个几何特征的偏差；第一区域中的几何特征中与第一几何特征V i的偏差最小的几何特征为与第一几何特征V i相匹配的第二几何特征U i，其中，i＝1,2，…，N。
其中，对于几何特征中的向量来说，第一几何特征V i与第二几何特征U i的偏差可以是第一几何特征V i中的向量与第二几何特征U i的中的向量之间的夹角。对于几何特征中的坐标点来说，第一几何特征V i与第二几何特征U i的偏差可以是第一几何特征V i中的坐标点与第二几何特征U i的中的坐标点之间的距离。
应理解，对于表面为曲线或曲面的对象来说，其几何特征可以包括多个向量和与向量一一对应的坐标点的集合，或者包括多个坐标点的集合，此时，在计算两个曲线的几何特征之间的偏差或两个曲面的几何特征之间的偏差时，应以多个包括坐标点的向量的或多个坐标点作为整体来计算。
应理解，当几何特征既包括向量也包括坐标点时，匹配的过程中可以仅考虑向量之间的夹角，也可以仅考虑两个坐标点之间的距离，也可以同时考虑上述两个量。
实现方式二：
为减少计算，加快匹配的过程，定位设备可以通过几何特征的属性来实现匹配。具体实现中，定位设备针对第一几何特征V i，从第一区域的几何特征(即M个几何特征)中选择与第一几何特征V i的属性相匹配的几何特征为与第一几何特征V i相匹配的第二几何特征U i，其中，i＝1,2，…，N。
其中，两个几何特征的属性相匹配包括两个几何特征的尺寸的误差小于第一阈值，两个几何特征所属对象的类别相同等，例如，两个几何特征的长度的误差小于0.01m；又例如，两个几何特征所属对象的类别都交通杆等。
实现方式三：
定位设备可以通过几何特征的属性和几何特征的结合来实现匹配。在具体的实现中，定位设备针对第一几何特征V i，可以从第一区域的几何特征(即M个几何特征)中选择与第一几何特征V i的属性相匹配的几何特征，若被选择的几何特征为多个时，则可以进一步地通过其他方式从被选择的几何特征中确定一个与第一几何特征V i相匹配的第二几何特征U i。例如，计算第一几何特征V i与被选择的几何特征中每一个几何特征的偏差，选择被选择的几何特征中与第一几何特征V i的偏差最小的几何特征为与第一几何特征V i相匹配的第二几何特征U i，其中，i＝1,2，…，N。
其中，两个几何特征的属性相匹配和两个几何特征的偏差的相关描述可以分别参见上述实现方式一和实现方式二中相关描述，此处不再赘述。
需要说明的是，定位设备也可以先选择第一区域中的几何特征中与第一几何特征V i的偏差小于第二阈值的几何特征，若被选择的几何特征为多个时，则可以进一步地比较两个几何特征的属性是否匹配，以从被选择的几何特征中确定一个与第一几何特征V i相匹配的第二几何特征U i。具体实现可以参照上述实现方式三中相关描述，此处不再赘述。
还需要说明的是，定位设备可以根据第一几何特征所属类别的不同选择不同的匹配实现方式。
例如，对于所属类别为树干、交通杆、建筑外缘等直线类型的第一几何特征或对于所述类别为建筑平面、墙面、柱平面，可以计算该第一几何特征的向量分别与第一区域中的几何特征中的每一个几何特征的向量的夹角，进而，第一区域中的几何特征中与该第一几何特征夹角最小的几何特征为该第一几何特征相匹配的第二几何特征。
应理解，定位设备实时获取到第一位置，并实时加载第一区域内的几何特征，并将第一区域内的几何特征与N个第一几何特征进行匹配。
本申请实施例中以N个第一几何特征为例来说明，N个第一几何特征可以定位设备从点云采集装置采集到的第一点云数据中提取得到的所有几何特征中选取出的N个第一几何特征；也可以是从点云采集装置采集到的第一点云数据中提取得到的所有几何特征。可以假设N个第一几何特征的集合表示为V＝{V 1,V 2,…V N}，几何特征地图上第二几何特征的集合表示为U＝{U 1,U 2,…U N}，其中，第一几何特征V i与第二几何特征U i相对应。S76具体的实现可以包括但不限于如下实施例(四)、实施例(五)和实施例(六)所描述五种定位方式。
实施例(四)
请参阅图9A所示的本申请实施例的一种定位方法的流程示意图。该定位方法中，S76的一种实现方式可以包括如下步骤：
S761：定位设备根据N个第一几何特征和几何特征地图中的N个第二几何特征确定几何特征之间的变换关系。
S762：根据几何特征之间的变换关系对车辆的第一位姿进行调整以得到第二位姿。
上述方法，通过观测到的第一几何特征和几何特征地图中的第二几何特征之间的变换关系对精确度低的第一位姿进行纠正，可得到精确度高的第二位姿，相对于现有技术中的点云数据，本申请实施例采用数据量少的几何特征进配准和定位，大大减小了数据的运算 量，使得车辆定位的耗时少，定位的实时性好。
第一定位方式：
第一定位方式的定位原理是：通过基于车辆当前的第一位姿所确定的第一坐标系与几何特征地图所采用的第二坐标系之间的变换关系(即本申请实施例中也称目标变换量)对车辆的第一位姿(即估计位姿)进行变换，得到车辆的第二位姿，即为车辆准确的位姿。其中，第一坐标系与第二坐标系之间的目标转换量即第一坐标系转换为第二坐标系需要进行旋转和平移，该旋转和平移可以通过计算第一几何特征与该第一几何特征对应的第二几何特征之间的旋转和平移得到。由于，针对不同的第一几何特征，第一几何特征与该第一几何特征对应的第二几何特征之间的旋转和平移可能不同。因此，为更精确地对车辆进行定位，定位设备可以将采集到的点云数据中提取得到的N个第一几何特征分别通过旋转R和平移t之后得到N个第三几何特征，使得该N个第三几何特征与N个第二几何特征之间的误差(即第一误差)最小的旋转R和平移t组成的转换量即为目标转换量的逆，N为大于1的正整数。如图9B所示，该第一定位方式的具体实现方式可以包括但不限于如下部分或全部步骤：
S911：初始化第一变换量。
S912：通过第一变换量对N个第一几何特征进行变换，得到N个第三几何特征，该第三几何特征与第一几何特征一一对应。
S913：针对第一几何特征V i，计算第一几何特征V i对应的第三几何特征与第一几何特征V i对应第二几何特征之间的误差，得到N个误差，i＝1,2,…，N。
S914：对N个误差进行求和，得到第一误差，其中，第一误差是对N个误差直接求和或加权求和得到。
S915：判断迭代次数或第一误差是否满足停止迭代条件。
在S915的一种具体实现中，定位设备可以判断迭代次数是否等于预设次数，该预设次数可以是4次、5次、10次、30次或其他数值，如果是，则执行S916；否则，执行S917。
在S915的另一种具体实现中，定位设备可以判断第一误差是否收敛，如果是，则执行S916；否则，执行S917。
应理解，本申请实施例还可以包括其他实现方式，例如，判断第一误差是否小于预设值，如0.1、0.2或其他数值等，对此不作限定。
S916：根据第一误差调节第一变换量。
在S916之后，定位设备重复执行S912-S915，直到迭代次数或第一误差满足停止迭代条件。
S917：输出的第一变换量T，第一变换量即为即为第一目标变换量。
S918：通过第一目标变换量T对第一位姿S1进行变换，得到车辆的第二位姿S2。可通过数学关系表达为：S2＝TS1，其中，
Figure PCTCN2019105810-appb-000015
其中，“由第一几何特征对应的第三几何特征与第一几何特征对应的第二几何特征之间的误差(即第一误差)”可以通过第一目标函数来计算，第一目标函数的输出值越高表示误差越大，调节初始变换量的过程就是尽可能缩小这个第一误差的过程。应理解，在第一变 换量的第一次调节过程中，第一变换量为定位系统预先设定的初始化的旋转量和平移量，通过多次迭代和调节，使得第一误差越来越小，使得第一误差最小的变换量即为第一目标变换量。还应理解，在本申请实施例的另一种实现中，第一目标变换量也可以是通过预设次数的调节得到的变换量。
在本申请实施例的一种实现中，可以构造第一目标函数，该第一目标函数可以包括但不限于以下3种形式：
第一目标函数的第一种形式：
定位设备可以根据第一目标函数确定第一误差，其中，第一目标函数可以是：
Figure PCTCN2019105810-appb-000016
ε为第一误差；第一变换量包括旋转R和平移t；w i为第一几何特征V i的权重；U i为第一几何特征V i对应的第二几何特征；i为N个第一几何特征中第一几何特征的索引，i为正整数，i≤N。
第一目标函数的第二种形式：
在一种实现中，第一几何特征可以仅包括向量，如直线的方向向量，平面的法向量、曲线的法向量曲面的法向量等，此时，第一几何特征V i＝v i，v i为第一几何特征V i的向量。此时，第一目标函数可以表示为：
Figure PCTCN2019105810-appb-000017
其中，旋转R和平移t即为第一变换量，为变量；i为N个第一几何特征中第一几何特征的索引，i为正整数，i≤N。本申请实施例中通过调节旋转R和平移t，不断地计算第一目标函数的值，使得第一目标函数最小的旋转R和平移t即为第一目标变换量。在本身实施例的另一种实现中也可以通过因式分解等方法得到使得第一目标函数最小的旋转R和平移t，本申请实施例不作限定。本申请实施例，(Rv i+t)即为第二几何特征U i的向量u i通过初始变换量进行变换得到的第三几何特征。
在上述第一目标函数的第一、二种形式中，w i为第一几何特征V i的权重，用于限制第一几何特征V i对第一目标变换量的贡献。其中，第一几何特征的权重可以根据第一几何特征相对于车辆或车辆上的点云采集装置的距离来确定，例如，第一几何特征V i的权重w i与第一几何特征V i所属的对象相对于车辆的距离负相关，即距离点云采集装置或车辆越近的第一几何特征具有越大的权重。
在上述第一目标函数的第一、二种形式中，第一几何特征的权重还可以根据第一几何特征所对应的对象的类型确定，即针对不同的对象可以设定不同的权重，例如，对于建筑平面来说，其得到的建筑平面的法向量精确度高，其对应的第一几何特征可以具有更高的权重；又例如，对于树干来说，其对应的直线的方向向量的精确度较低，其对应的第一几何特征可以具有较低的权重。应理解，不限于上述权重的设置方式，本申请实施例还可以通过其他设置方式来设置权重，比如，综合考虑第一几何特征的距离和第一几何特征对应的对象的类型；又例如，本申请实施例中还可以不包括权重，即针对任意一个第一几何特征，其权重为1，此处不作限定。
第一目标函数的第三种形式：
在一种实现中，第一几何特征可以包括向量和位置，此时，第一目标函数可以表示为：
Figure PCTCN2019105810-appb-000018
其中，G 1,i为第一几何特征V i的位置，G 2,i为第二几何特征U i的位置；w i为第一几何特征V i的向量v i的权重，具体可参见上述第一目标函数的第一种形的中相关描述，此处不再赘述；
Figure PCTCN2019105810-appb-000019
第一几何特征V i的位置G 1,i的权重，用于限定第一几何特征V i的位置G 1,i对第一目标变换量的贡献。其中，第一几何特征的位置的权重可以由对应的对象的类型确定，即针对不同的对象可以设定不同的权重，例如，交通杆来说，其位置按照位置的设定规则可以精确的确定，则其几何特征的位置具有较高的权重，比如1；又例如，对于路沿来说，很难定义精确的位置，则其几何特征的位置具有较低的权重，比如0。
应理解，不限于上述权重
Figure PCTCN2019105810-appb-000020
的设置方式，本申请实施例还可以通过其他设置方式来设置权重
Figure PCTCN2019105810-appb-000021
比如，综合考虑第一几何特征V i的距离和第一几何特征V i对应的对象的类型来设定
Figure PCTCN2019105810-appb-000022
对此，本申请实施例不作限定。
第二定位方式：
在第一中定位方式中，变换关系是以将第一几何特征变换为几何特征地图中的第二几何特征变换为例来说明，应理解，变换关系还可以是将几何特征地图中的第二几何特征变换为第一几何特征。
与第一定位方式的定位原理相似，不同的是变换关系的计算方法。在第二定位方式中，定位设备可以将N个第二几何特征分别通过旋转R和平移t之后得到N个第四几何特征，使得该N个第四几何特征与采集到的点云数据中提取得到的N个第一几何特征之间的误差最小的旋转R′和平移t′即为目标转换量T的逆矩阵，N为大于1的正整数。如图9C所示，该第二定位方式的具体实现方式可以包括但不限于如下部分或全部步骤：
S921：初始化第二变换量。
S922：通过第二变换量对多个第二几何特征进行变换，得到多个第四几何特征，该第四几何特征与第二几何特征一一对应。
S923：针对第二几何特征U i，计算第二几何特征U i对应的第四几何特征与第二几何特征U i对应第一几何特征之间的误差，得到N个误差，i＝1,2,…，N。
S924：对N个误差进行求和，得到第二误差，其中，第二误差是对N个误差直接求和或加权求和得到。
S925：判断迭代次数或第二误差是否满足停止迭代条件。具体实现可以参见上述S914中相关描述，在判断结果为是，则执行S926；否则，执行S927。
S926：根据第二误差调节第二变换量。
在S926之后，定位设备重复执行S922-S925，直到迭代次数或误差满足停止迭代条件。
S927：输出的第二变换量，输出的第二变换量的逆矩阵即为即为第二目标变换量。
S928：通过第二目标变换量T′对第一位姿S1进行变换，得到车辆的第二位姿S2。可 通过数学关系表达为：S2＝T′S1，其中，
Figure PCTCN2019105810-appb-000023
上述S921-S928的具体实现可以参见上述第一定位方式中相关描述，此处不再赘述。
同理，可以构造第二目标函数来实现第二误差的计算。同第一目标函数的第一种形式相似，第二目标函数可以表示为：
Figure PCTCN2019105810-appb-000024
同第一目标函数的第二种形式相似，第二目标函数可以表示为：
Figure PCTCN2019105810-appb-000025
关于，w i、
Figure PCTCN2019105810-appb-000026
u i、v i、G 1,i、G 2,i、i的相关描述可以参见上述第一定位方式中相关描述，此处不再赘述。
应理解，不限于上述第一目标函数或第二目标函数的两种形式中，第一目标函数或第二目标函数还可包括其他形式，误差还可以是平均绝对误差(mean absolute error，MAE)、均方误差(mean squared error，MSE)、均方根误差(root mean squared error，RMSE)或其它的形式等，本申请不作限定。
还应理解，目标转换量的计算过程即为第一目标函数或第二目标函数最小化的求解，目标函数最小化的求解方法可以包括但不限于高斯牛顿法、梯度下降法、L-M(Levenberg-Marquardt)法、QR分解或其他求解方法。
实施例(五)
请参阅图10A所示的定位方法的流程示意图，该方法除包括上述S72、S74、S76之外还可以包括S71、S75，具体可参见上述实施例(三)中相关描述，其中S76可以包括但不限于如下步骤：
S763：根据上一时刻的第二位姿和上一时刻时车辆的控制参数，确定车辆在当前时刻的预测位姿，所述上一时刻为当前时刻之前的时刻。
S764：通过第二参数的观测值与第二参数的预测值之间的误差更新车辆的预测位姿，得到车辆的第二位姿。
其中，第二参数的观测值是基于车辆在第一位姿下观测到的第一几何特征确定的，第二参数的预测值是基于预测位姿和几何特征地图中的第二几何特征确定的。
应理解，上述第一位姿可以是当前时刻的预测位姿，也可以是通过其他方法定位到的车辆当前时刻的位姿，比如，定位设备根据定位系统确定车辆的第一位置以及根据惯性传感器确定车辆的第一姿态，第一位姿包括第一位置和第一姿态。
在本申请实施例的一种实现中，在S764之前，该定位设备还可以获取车辆观测到的N个第一几何特征，一个第一几何特征用于在由第一位姿确定的空间中表示车辆观测到的一个对象的几何特征，N为正整数；在几何特征地图中查找与N个第一几何特征相匹配的N个第二几何特征，第一几何特征与第二几何特征一一对应；进而，基于所述第一位姿、预测位姿、N个第一几何特征和N个第二几何特征确定第二参数的观测值与第二参数的预测 值之间的误差。
应理解，定位设备获取N个第一几何特征的具体实现可以分别参见上述实施例(三)中相关描述；定位设备在几何特征地图中查找与N个第一几何特征相匹配的N个第二几何特征的具体实现可以分别参见上述实施例(三)中步骤S75以及匹配过程的三种实现方式的相关描述，此处不再赘述。
实施例(五)提供了一种车辆定位方法，通过观测到的第一几何特征和几何特征地图中的第二几何特征对估计位姿进行评分，确定评分最高的估计位姿作为车辆的实际位姿，相对于现有技术中的点云数据，本申请实施例采用数据量少的几何特征进行定位，大大减小了数据的运算量，使得车辆定位的耗时少，定位的实时性好。
步骤S76具体可以通过Kalman滤波的方法来计算车辆的第二位姿。如图10B所示第三定位方式的流程示意图和图10C所示的第三种定位方式的示意性说明图，第三定位的实现方式包括Kalman预测过程和Kalman更新过程，第三定位的实现方式包括但不限于如下步骤：
S102：将车辆在上一时刻(k-1)的第二位姿
Figure PCTCN2019105810-appb-000027
和上一时刻时车辆的控制输入a k输入到车辆的运动学方程，预测车辆在当前时刻(k)的预测位姿
Figure PCTCN2019105810-appb-000028
其中，k为大于1的正整数。
应理解，S102为Kalman预测过程，可以是上述S71的一种实现方式，此时第一位姿即为预测位姿。在本申请实施例中，可以根据车辆的动力学方程(即本申请实施例钟红运动方程)得到车辆的状态。本申请实施例中，车辆的状态为位姿，包括位置和姿态。在本申请的一种实现中，以
Figure PCTCN2019105810-appb-000029
为例来说明，本申请实施例以2维的坐标
Figure PCTCN2019105810-appb-000030
来表示车辆的位置，以偏航角
Figure PCTCN2019105810-appb-000031
来表示车辆的偏航角。在本申请的另一种实现中，也可以通过3维的坐标来表示车辆的位置，以3个角度(俯仰角、偏航角和翻滚角)来表示车辆的姿态，对此本申请实施例不作限定。通过上述车辆的运动方程进行位姿预测。在车辆的运动方程(也即为本申请实施例中的预测方程)可以表示为：
Figure PCTCN2019105810-appb-000032
其中，A k为当前时刻下的状态转移矩阵；B k为当前时刻下的控制输入矩阵；a k为车辆的控制输入，如加速度、转向等；ω k为状态的噪声，其均值为0，协方差矩阵为Q k。
当前时刻(k)协方差矩阵
Figure PCTCN2019105810-appb-000033
根据上一时刻(k-1)的协方差矩阵
Figure PCTCN2019105810-appb-000034
来进行预测：
Figure PCTCN2019105810-appb-000035
S104：根据预测位姿
Figure PCTCN2019105810-appb-000036
和N个第二几何特征确定第二参数的预测值
Figure PCTCN2019105810-appb-000037
具体地，可以将预测位姿
Figure PCTCN2019105810-appb-000038
和N个第二几何特征输入到观测方程，得到当前时刻(k) 的第二参数的预测值。
其中，观测方程基于第二参数的方程，第二参数可以是第一几何特征所描述对象相对于自身车辆距离、方位角和高度角等中的至少一种。第二参数的预测值为N个第二几何特征分别相对于预测位姿下的车辆的第二参数。第二参数的观测值为N个第一几何特征分别相对于在第一位姿下的车辆的第二参数。本申请实施例以第二参数为第一几何特征描述的对象相对于车辆的距离和方位角为例来说明，应理解，还可以应用Kalman滤波构造其他的第二参数和观测方程。应理解，第一几何特征所描述的对象与该第一几何特征对应的第二几何特征所描述的对象为同一对象。
对于通过位置点来描述的几何特征来说，第二参数可以是该位置点相对于车辆自身的距离和方位角。其中，若N个第一几何特征包括对象的位置，或者通过多个坐标点来表示，当前时刻(k)的第二参数的预测值可以表示为：
Figure PCTCN2019105810-appb-000039
其中，第一几何特征V i相对于预测位姿下的车辆的第二参数，即第一几何特征V i对应的第二参数的预测值
Figure PCTCN2019105810-appb-000040
为：
Figure PCTCN2019105810-appb-000041
其中，G 2,i＝[x 2,i，y 2,i]表示第二几何特征U i中的坐标点，本申请实施例中以2维坐标为例来说明，应理解，在本申请另一种实现中，G 2,i还可以以3维坐标来表示，即G 2,i＝[x 2,i，y 2,i，z 2,i]，对此，本申请实施例不作限定。
对于通过向量来描述的几何特征来说，第二参数可以是向量相对于自身车辆的距离、方位角和高度角。若N个第一几何特征包括对象的方向，即通过方向向量来表示，当前时刻(k)的第二参数的预测值可以表示为，
Figure PCTCN2019105810-appb-000042
此时，第一几何特征V i相对于预测位姿下的车辆的第二参数，即第一几何特征V i对应的第二参数的预测值
Figure PCTCN2019105810-appb-000043
为：
Figure PCTCN2019105810-appb-000044
其中，
Figure PCTCN2019105810-appb-000045
为当前时刻的预测距离，
Figure PCTCN2019105810-appb-000046
为当前时刻预测方位角，
Figure PCTCN2019105810-appb-000047
为当前时刻的预测高度角。请参阅图10D所示的向量的高度角和方位角的示意图。其中，车辆坐标系的原点到向量的距离OA1即为
Figure PCTCN2019105810-appb-000048
坐标A1到车辆坐标系OXY平面的投影为坐标A2，则，OA2与OX的夹角即为
Figure PCTCN2019105810-appb-000049
OA2与OA1的夹角即为
Figure PCTCN2019105810-appb-000050
S106：根据第一位姿和N个第一几何特征确定
第二参数的观测值。
具体地，定位设备可以将第一位姿S1和N个第一几何特征输入到观测方程，得到当前时刻(k)第二参数的观测值。
对应于上述公式(6)中的第二参数，当前时刻(k)的第二参数的观测值可以表示为Z k＝[Z k,1 T Z k,2 T ... Z k,N T] T，其中，第一几何特征V i相对于第一位姿下的车辆的第二参数，即第一几何特征V i对应的第二参数的观测值Z k,i为：
Figure PCTCN2019105810-appb-000051
对应于车辆运动方程中位姿的表达形式，第一位姿可以表示为
Figure PCTCN2019105810-appb-000052
(x int,y int)表示车辆的第一位置，θ int表示车辆的第一姿态。
S108：根据第二参数的观测值
Figure PCTCN2019105810-appb-000053
和第二参数的预测值Z k确定第二参数的观测值
Figure PCTCN2019105810-appb-000054
与第二参数的预测值Z k之间的误差E k。
其中，N个第一几何特征在当前时刻的第二参数的预测值可以表示为
Figure PCTCN2019105810-appb-000055
N个第一几何特征的第二参数的观测值可以表示为Z k＝[Z k,1 T Z k,1 T ... Z k,1 T] T，则当前时刻的第三误差可以表示为
Figure PCTCN2019105810-appb-000056
S110：根据第二参数的观测值
Figure PCTCN2019105810-appb-000057
与第二参数的预测值Z k之间的误差E k，更新当前时刻的预测位姿
Figure PCTCN2019105810-appb-000058
得到车辆在当前时刻的第二位姿
Figure PCTCN2019105810-appb-000059
应理解，S110为Kalman更新过程。更新使得状态误差最小，可以得到位姿更新的方程为：
Figure PCTCN2019105810-appb-000060
卡尔曼增益的更新方程为：
Figure PCTCN2019105810-appb-000061
位姿的协方差矩阵的更新方程为：
Figure PCTCN2019105810-appb-000062
其中，R k、Q k分别为观测噪声矩阵和位姿噪声矩阵。A k、H k分别为车辆位姿转换矩阵和观测转换矩阵。更新后得到的
Figure PCTCN2019105810-appb-000063
即为当前时刻的第二位姿。
在步骤S110之后，定位设备可以将当前时刻的第二位姿
Figure PCTCN2019105810-appb-000064
输入到运动方程、预测出下一时刻(k+1)的预测位姿
Figure PCTCN2019105810-appb-000065
第二位姿，进而，通过上述S71-S76或S102-S110所述的方法，得到车辆在下一时刻(k+1)的第二位姿
Figure PCTCN2019105810-appb-000066
实施例(六)
请参阅图11A所示的定位方法的流程示意图，该方法除包括上述S72、S74、S76之外还可以包括S71、S75，其中S76可以包括但不限于如下步骤：
S705：根据第一位姿，对车辆的位姿进行估计，得到多组估计位姿。
S706：根据N个第一几何特征和几何特征地图中的N个第二几何特征确定多组估计位姿的评分。
S707：根据多组估计位姿中每一组估计位姿的评分确定车辆的第二位姿，其中，其中，第一组估计位姿的评分用于指示第一组估计位姿与第二位姿的接近程度，第一组估计位姿是多组估计位姿中任意一个估计位姿。
可选地，第一组估计位姿的评分是基于第一组估计位姿、第一位姿、车辆观测到的N个第一几何特征和几何特征地图中的N个第二几何特征确定的。
在本申请实施例的一种实现中，在S706之前，该定位设备还可以获取车辆观测到的N个第一几何特征，一个第一几何特征用于在由第一位姿确定的空间中表示车辆观测到的一个对象的几何特征，N为正整数；进而，在几何特征地图中查找与N个第一几何特征相匹配的N个第二几何特征，第一几何特征与第二几何特征一一对应；进而，基于每一组估计位姿、第一位姿、N个第一几何特征、N个第二几何特征确定每一组估计位姿的评分。
应理解，定位设备获取车辆观测到的N个第一几何特征的具体实现可以分别参见上述实施例(三)中相关描述；定位设备在几何特征地图中查找与N个第一几何特征相匹配的N个第二几何特征的具体实现可以分别参见上述实施例(三)中步骤S75以及匹配过程的三种实现方式的相关描述，此处不再赘述。
实施例(六)中，通过观测到的第一几何特征确定的第二参数的观测值和基于几何特征地图中的第二几何特征确定的第二参数的预测值之间的误差对当前时刻的预测位姿进行更新，得到车辆的实际位姿，相对于现有技术中的点云数据，本申请实施例采用数据量少的几何特征进行定位，大大减小了数据的运算量，使得车辆定位的耗时少，可实现车辆实时定位。
在本申请实施例中，定位设备对第一位姿进行调整得到第二位姿至少可以通过如下三种定位方式实现：
第四定位方式：
在第四定位方式中，可以对车辆的位姿进行先验位姿估计，得到多组估计位姿和在每一组估计位姿时的第一参数的估计值；进而，针对每一个估计位姿，根据第一参数的估计值和第一参数的观测值的误差得到第一评分，通过第一评分评价其与第二位姿的接近程度。应理解，估计位姿对应的估计值和观测值的误差越小，则该估计位姿的第一评分越高，该估计位姿越接近车辆的实际位姿，即第二位姿。如图11B所示为本申请实施例提供的第四定位方式的流程示意图，该第四定位方式可以包括但不限于如下步骤：
S11011：根据第一位姿S1对车辆的第二位姿S2进行先验估计，得到D组估计位姿，D为大于1的正整数。
在S1102的一种实现中，第一位姿S1可以是定位设备根据上一时刻的第二位姿和车辆的运动方程预测得到的当前时刻的预测位姿，即为上述第三定位方式中描述的
Figure PCTCN2019105810-appb-000067
可参见上述第三定位方式中相关描述，此处不再赘述。应理解，第一位姿还可以是通过GPS和车 辆的惯性传感器定位得到的当前时刻的位姿，或其他方式定位得到的位姿，本申请实施例不作限定。
D组估计位姿可以是以第一位姿S1为期望，在第一位姿S1周围以正态分布的位姿的集合。D可以为100、1000、3000或其他数值等。
本申请实施例中，D组估计位姿可以表示为
Figure PCTCN2019105810-appb-000068
S11012：根据每一组估计位姿和N个第二几何特征，确定每一组估计位姿对应的第一参数的估计值。
本申请实施例中以估计位姿
Figure PCTCN2019105810-appb-000069
为例来进行说明，定位设备可以将估计位姿
Figure PCTCN2019105810-appb-000070
和N个第二几何特征输入到观测方程，得到估计位姿
Figure PCTCN2019105810-appb-000071
对应的第一参数的估计值
Figure PCTCN2019105810-appb-000072
估计位姿
Figure PCTCN2019105810-appb-000073
为该D组估计位姿中的一个估计位姿。其中，估计位姿
Figure PCTCN2019105810-appb-000074
对应的第一参数的估计值可以表示为：
Figure PCTCN2019105810-appb-000075
其中，j为D组估计位姿中估计位姿的索引，j＝1,2,…，D。其中，
Figure PCTCN2019105810-appb-000076
是将估计位姿
Figure PCTCN2019105810-appb-000077
和第二几何特征U i输入到观测方程得到的第一参数的估计值。
可选地，同上述实施例(五)中第二参数相似，观测方程基于第一参数的方程，第一参数可以是第一几何特征所描述对象相对于自身车辆距离、方位角和高度角等中的至少一种。第一参数的估计值为N个第二几何特征分别相对于估计位姿下的车辆的第一参数。第一参数的观测值为N个第一几何特征分别相对于在第一位姿下的车辆的第一参数。
关于车辆的运动学方程、观测方程的相关描述可以参见上述实施例(五)中相关描述，此处不再赘述。关于估计位姿对应的第一参数的估计值的计算方式可以参见上述实施例(五)中第二参数的预测值的计算方式(即步骤S104)中相关描述，此处不再赘述。
可以理解，通过上述S11012可以得到D组估计位姿分别对应的第一参数的估计值。此时，D组估计位姿对应的第一参数的估计值可以表示为：
Figure PCTCN2019105810-appb-000078
S11013：根据第一位姿和N个第一几何特征，确定第一参数的观测值。
将第一位姿和N个第一几何特征输入到观测方程，得到第一参数的观测值。
同理，关于第一参数的观测值的计算方式可以参见上述实施例(五)中第二参数的观测值的计算方式(即步骤S106)中相关描述，此处不再赘述
S11014：根据每一组估计位姿对应的第一参数的估计值与第一参数的观测值之间的误差确定每一组估计位姿的第一评分。
应理解，通过上述S1104中估计位姿的评分在第四种定位方式中也称为估计位姿的第一评分。
可以理解，根据估计位姿对应的第一参数的估计值与第一参数的观测值之间的误差越小，说明该估计位姿越接近与第二位姿(实际位姿)。若估计位姿对应的第一参数的估计值与第一参数的观测值相等，则该估计位姿就是第二位姿。可以根据估计位姿对应的第一参 数的估计值与第一参数的观测值之间的误差构造用于计算第一评分的函数。第一评分越高，则该第一评分对应的估计位姿与第二位姿越接近。例如，估计位姿
Figure PCTCN2019105810-appb-000079
的第一评分可以通过下式计算：
Figure PCTCN2019105810-appb-000080
其中，corr()为皮尔逊积距相关系数(Pearson product-moment correlation coefficient)，可以是估计位姿
Figure PCTCN2019105810-appb-000081
对应的第一参数的估计值
Figure PCTCN2019105810-appb-000082
与第二参数的观测值Z k的协方差与二者标准差积的商。
应理解，通过上述S11011-S11014所述的方法，可以得到D组估计位姿中每一个估计位姿的第一评分。
S11015：根据D组估计位姿和该D组估计位姿中每一组估计位姿的第一评分，得到第二位姿。
在S11015的一种实现中，第二位姿可以是最高的D组估计位姿中第一评分最高的估计位姿。
在S11015的另一种实现中，第二位姿S2可以表示为：
Figure PCTCN2019105810-appb-000083
其中，μ j为score j的归一化系数。
应理解，还可以基于D组估计位姿和该D组估计位姿中每一组估计位姿的评分，通过其他方式得到第二位姿，本申请实施例不作限定。
第五定位方式：
在第四定位方式中，若第一参数为第二几何特征本身，即为第五定位方式所描述的定位方式。如图11C所示为本申请实施例提供的第五定位方式的流程示意图，该第五定位方式可以包括但不限于如下步骤：
S11021：根据第一位姿S1对车辆的第二位姿S2进行先验估计，得到D组估计位姿，D为大于1的正整数。
关于S11021的具体实现可以参见上述第四定位方式中S11011中相关描述，此处不再赘述。
S11022：通过每一组估计位姿与第一位姿之间的变换关系对N个第二几何特征分别进行变换得到的每一组估计位姿对应的N个第五几何特征，N个第二几何特征与N个第五几何特征一一对应。
本申请实施例中，以估计位姿
Figure PCTCN2019105810-appb-000084
为例，通过第一估计位姿和第一位姿的变换关系对N个第二几何特征分别进行变换，得到该第一估计位姿对应的N个第五几何特征，估计位姿
Figure PCTCN2019105810-appb-000085
为D组估计位姿中的一个估计位姿，j为D组估计位姿中估计位姿的索引，j＝1,2,…，D。第二几何特征与第五几何特征一一对应。
应理解，第二几何特征和第五几何特征是针对同一对象在不同坐标系下的表达。若估计位姿
Figure PCTCN2019105810-appb-000086
为车辆的第二位姿(实际位姿)，理论上，该估计位姿
Figure PCTCN2019105810-appb-000087
对应的N个第五几何特征与N个第一几何特征相同。也就是说，若估计位姿
Figure PCTCN2019105810-appb-000088
对应的N个第五几何特征与N个第一几何特征存在误差可以评估估计位姿
Figure PCTCN2019105810-appb-000089
与第二位姿的接近程度，即得到估计位姿
Figure PCTCN2019105810-appb-000090
的评分(本申请实施例中也称第二评分)。该第二评分用于评价估计位姿与第二位姿的接近程度。第二评分越高，则该第二评分对应的估计位姿与第二位姿越接近。
S11023：根据每一组估计位姿对应的N个第五几何特征与N个第一几何特征之间的误差确定每一组估计位姿的第二评分。
估计位姿
Figure PCTCN2019105810-appb-000091
对应的N个第五几何特征可以表示为 jO＝[ jO 1  jO 2 …  jO N]，估计位姿
Figure PCTCN2019105810-appb-000092
对应的N个第五几何特征与N个第一几何特征V＝{V 1,V 2,…V N}的误差可以表示为：
Figure PCTCN2019105810-appb-000093
其中，N个第一几何特征中N 1个第一几何特征通过向量表示、N 2个第一几何特征通过坐标点表示，m为N 1个第一几何特征中第一几何特征的索引，n为N 2个第一几何特征中第一几何特征的索引，m、n、N 1、N 2为正整数，且m≤N 1、n≤N 2、N 1+N 2＝N； αw m表示第一几何特征U m的权重、 dw n表示第一几何特征U n的权重，同公式(2)中类似，第一几何特征的权重的可以根据该第一几何特征与车辆的距离确定，具体可以参见上述第一种定位方式中相关描述，此处不再赘述；<V m, jO m>表示两个向量的夹角，对应的权重系数为 αw m；特征为目标的位置时，||V n- jO n|| 2表示两者的距离，对应的权重系数为 dw n。
此时，估计位姿
Figure PCTCN2019105810-appb-000094
的评分可以表示为：
Figure PCTCN2019105810-appb-000095
通过上述S11022和S11023，可以得到D组估计位姿中每一组估计位姿的评分。
S11024：根据D组估计位姿和D组估计位姿中每一组估计位姿的第二评分，得到第二位姿。
其中，S11024的具体实现可参见同上述第四定位方式中S11014中相关描述，此处不再赘述。
第六定位方式：
在第四定位方式中，若第一参数为第一几何特征本身，即为第六定位方式所描述的定位方式。如图11D所示为本申请实施例提供的第六定位方式的流程示意图，该第六定位方式可以包括但不限于如下步骤：
S11031：根据第一位姿S1对车辆的第二位姿S2进行先验估计，得到D组估计位姿，D为大于1的正整数。
关于S11031的具体实现可以参见上述第四定位方式中S11011中相关描述，此处不再赘述。
S11032：通过每一组估计位姿与第一位姿之间的变换关系对N个第一几何特征分别进 行变换得到的每一组估计位姿对应的N个第六几何特征，N个第一几何特征与N个第六几何特征一一对应。
应理解，第一几何特征和第六几何特征是针对同一对象在不同坐标系下的表达。若估计位姿
Figure PCTCN2019105810-appb-000096
为车辆的第二位姿(实际位姿)，理论上，该估计位姿
Figure PCTCN2019105810-appb-000097
对应的N个第五几何特征与N个第二几何特征相同。也就是说，若估计位姿
Figure PCTCN2019105810-appb-000098
对应的N个第六几何特征与N个第二几何特征存在误差可以评估估计位姿
Figure PCTCN2019105810-appb-000099
与第二位姿的接近程度，即得到估计位姿
Figure PCTCN2019105810-appb-000100
的评分(也称为第三评分)。该第三评分用于评价估计位姿与第二位姿的接近程度。第三评分越高，则该第三评分对应的估计位姿与第二位姿越接近。
S11033：根据每一组估计位姿对应的N个第六几何特征与N个第二几何特征之间的误差确定每一组估计位姿的第三评分。
应理解，S11033中估计位姿
Figure PCTCN2019105810-appb-000101
对应的N个第六几何特征与N个第二几何特征之间的误差的计算方式可以参照上述第五种定位方式中估计位姿
Figure PCTCN2019105810-appb-000102
对应的N个第五几何特征与N个第一几何特征之间的误差的计算方式(即步骤S11023)中相关描述，此处不再赘述。
S10134：根据D组估计位姿和D组估计位姿中每一组估计位姿的第三评分，得到第二位姿。
应理解，S10134的具体实现可参照上述第五种定位方式中步骤S11024中相关描述，此处不再赘述。
需要说明的是，上述定位方法可以应用于车辆的导航、车辆的自动驾驶等需要对车辆的位姿进行精确估计的场景中，通过上述定位方法为不仅可以得到车辆提供更精准位置还可以得到车辆的姿态。
下面介绍本申请实施例涉及的装置、设备。
图12为本发明实施例中一种定位装置的示意性框图，图12所示的定位装置(该装置1200具体可以是图2对应实施例中的定位设备，比如车辆160/终端180/定位服务器190，也可以是图3中车辆100，具体地，该定位装置1200可以包括：
第一获取单元1201，用于车辆通过点云采集装置采集的第一点云数据；
特征提取单元1202，用于从该第一点云数据中提取N个第一几何特征，N为正整数；
调整单元1203，用于根据N个第一几何特征对该车辆的第一位姿进行调整以得到该车辆的第二位姿，第二位姿的精确度高于第一位姿的精确度。
在本申请一种可能的实现中，该装置1200还包括：
第二获取单元1204，用于在调整单元根据所述N个第一几何特征对所述车辆的第一位姿进行调整以得到所述车辆的第二位姿之前，获取该车辆的第一位姿。
在本申请一种可能的实现中，该装置1200还包括：
匹配单元1205，用于在几何特征地图中查找与该N个第一几何特征相匹配的N个第二几何特征。
在本申请实施例的另一种定位装置中，该定位装置可以不包括上述定位装置1200中的第一获取单元1201和特征提取单元1202，定位装置1200中调整可以包括接收单元，用于接收车辆或终端发送的N个第一几何特征。
需要说明的是，定位装置1200中第二获取单元1204、匹配单元1205不是定位装置1200必须的单元。还需要说明的是，上述定位装置1200还包括其他用于实现实施例(三)、实施例(四)、实施例(五)或实施例(六)所述的定位方法中的单元，上述定位装置1200的各个单元或其他单元的具体实现以参见上述实施例(三)、实施例(四)、实施例(五)或实施例(六)中相关描述，此处不再赘述。
图13是本申请实施例提供的一种定位装置的硬件结构示意图。图13所示的定位装置1300(该装置1300具体可以是一种计算机设备)包括存储器1301、处理器1302、通信接口1303以及总线1304。其中，存储器1301、处理器1302、通信接口1303通过总线1304实现彼此之间的通信连接。
存储器1301可以是只读存储器(Read Only Memory，ROM)，静态存储设备，动态存储设备或者随机存取存储器(Random Access Memory，RAM)。存储器1301可以存储程序，当存储器1301中存储的程序被处理器1302执行时，处理器1302和通信接口1303用于执行本申请实施例(三)的定位方法的各个步骤。
处理器1302可以采用通用的中央处理器(Central Processing Unit，CPU)，微处理器，应用专用集成电路(Application Specific Integrated Circuit，ASIC)，图形处理器(graphics processing unit，GPU)或者一个或多个集成电路，用于执行相关程序，以实现本申请实施例的定位装置1200中的单元所需执行的功能，或者执行本申请方法本申请实施例(三)的定位方法。
处理器1302还可以是一种集成电路芯片，具有信号的处理能力。在实现过程中，本申请的样本生成方法的各个步骤可以通过处理器1302中的硬件的集成逻辑电路或者软件形式的指令完成。上述的处理器1302还可以是通用处理器、数字信号处理器(Digital Signal Processing，DSP)、专用集成电路(ASIC)、现成可编程门阵列(Field Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为硬件译码处理器执行完成，或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器，闪存、只读存储器，可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器1301，处理器1302读取存储器1301中的信息，结合其硬件完成本申请实施例的定位装置1200中包括的单元所需执行的功能，或者执行本申请实施例(三)、实施例(四)、实施例(五)或实施例(六)中的定位方法。
通信接口1303使用例如但不限于收发器一类的收发装置，来实现装置1300与其他设备或通信网络之间的通信。例如，可以通过通信接口1303获取点云数据、第一几何特征、第二几何特征、几何特征地图等数据。通信接口1303还用于实现与其他设备，例如几何特征地图生成设备、地图服务器、终端等之间的通信。
总线1304可包括在装置1300各个部件(例如，存储器1301、处理器1302、通信接口1303)之间传送信息的通路。
当定位装置1300设置于车辆上时，该车辆可以是上述图3示的车辆100，装置1300还可以包括点云采集装置1305，该定位装置1300还可以执行实施例(一)中的几何特征提取方法。点云采集装置1305可以是激光雷达(laser radar)，立体摄像头(stereo camera)，越渡时间相机(timeof flight camera)等可以获取点云数据的装置。
对于定位装置1300为终端，如手机、平板电脑，或为服务器、云端等时，点云采集装置1305不是其必须的器件。
应理解，定位装置1200中的第一获取单元1201、第二获取单元1204、接收单元可以相当于装置1300中的通信接口1303，提取单元1202、调整单元1203和匹配单元1205可以相当于处理器1302。
上述各个器件的具体实现可以参见上述实施例(三)、实施例(四)、实施例(五)或实施例(六)中相关描述，本申请实施例不再赘述。
图14为本发明实施例中一种几何特征的提取装置的示意性框图，图14所示的几何特征的提取装置(该装置1400具体可以是图2对应实施例中定位设备，比如车辆160/终端180/定位服务器190，或者该装置1400可以是几何特征地图生成设备140)，具体地，该装置1400可以包括：
获取单元1401，用于获取待处理的点云数据；
提取单元1402，用于从该待处理的点云数据中提取至少一个几何特征；其中，该至少一个几何特征用于车辆的定位。
可选地，装置1400还可以包括地图生成单元，用于根据提取得到的几何特征生成几何特征地图。
需要说明的是，装置1400还包括其他用于实现实施例(一)所述的几何特征提取方法中的单元或实施例(二)描述的几何特征地图生成方法中的单元，上述装置1400的各个单元或其他单元的具体实现以参见上述实施例(一)或实施例(二)中相关描述，此处不再赘述。
图15是本申请实施例提供的一种几何特征的提取装置的硬件结构示意图。图15所示的几何特征的提取装置1500(该装置1500具体可以是一种计算机设备)包括存储器1501、处理器1502、通信接口1503以及总线1504。其中，存储器1501、处理器1502、通信接口1503通过总线1504实现彼此之间的通信连接。
存储器1501可以是只读存储器(Read Only Memory，ROM)，静态存储设备，动态存储设备或者随机存取存储器(Random Access Memory，RAM)。存储器1501可以存储程序，当存储器1501中存储的程序被处理器1502执行时，处理器1502和通信接口1503用于执行本申请实施例(一)或实施例(二)所述的方法的各个步骤。
处理器1502可以采用通用的中央处理器(Central Processing Unit，CPU)，微处理器，应用专用集成电路(Application Specific Integrated Circuit，ASIC)，图形处理器(graphics processing unit，GPU)或者一个或多个集成电路，用于执行相关程序，以实现本申请实施例的定位装置1400中的单元所需执行的功能，或者执行本申请方法本申请实施例(一)或 实施例(二)所述的方法。
处理器1502还可以是一种集成电路芯片，具有信号的处理能力。在实现过程中，本申请的样本生成方法的各个步骤可以通过处理器1502中的硬件的集成逻辑电路或者软件形式的指令完成。上述的处理器1502还可以是通用处理器、数字信号处理器(Digital Signal Processing，DSP)、专用集成电路(ASIC)、现成可编程门阵列(Field Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为硬件译码处理器执行完成，或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器，闪存、只读存储器，可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器1501，处理器1502读取存储器1501中的信息，结合其硬件完成本申请实施例的定位装置1400中包括的单元所需执行的功能，或者执行本申请方法实施例(一)或实施例(二)所述的方法。
通信接口1503使用例如但不限于收发器一类的收发装置，来实现装置1500与其他设备或通信网络之间的通信。例如，可以通过通信接口1503获取点云数据、第一几何特征、第二几何特征、几何特征地图等数据。通信接口1503还用于实现与其他设备，例如几何特征地图生成设备、地图服务器、终端等之间的通信。
总线1504可包括在装置1500各个部件(例如，存储器1501、处理器1502、通信接口1503)之间传送信息的通路。
当装置1500设置于车辆上时，该车辆可以是上述图3示的车辆100，装置1500还可以包括点云采集装置1505，该定位装置1500还可以执行实施例(一)中的几何特征提取方法。点云采集装置1505可以是激光雷达(laser radar)，立体摄像头(stereo camera)，越渡时间相机(timeof flight camera)等可以获取点云数据的装置。
若装置1500为终端，如手机、平板电脑，或为服务器、云端等，点云采集装置1505不是其必须的器件。
应理解，装置1400中的获取单元1401可以相当于装置1500中的通信接口1503，提取单元1402可以相当于处理器1502。
上述各个器件的具体实现可以参见上述实施例(一)或实施例(二)中相关描述，本申请实施例不再赘述。
应注意，尽管图13和图15所示的装置1300、1500仅示出了存储器、处理器、通信接口或无线通信模块，但是在具体实现过程中，本领域的技术人员应当理解，装置1300、1500还包括实现正常运行所必须的其他器件。同时，根据具体需要，本领域的技术人员应当理解，装置1300、1500还可包括实现其他附加功能的硬件器件。此外，本领域的技术人员应当理解，装置1300、1500也可仅包括实现本申请实施例所必须的器件，而不必包括图13或图15中所示的全部器件。
例如，装置1300、1500还可以包括输入/输出装置，该输入装置可以是触控面板、麦克风或其他输出装置等。输出装置可以是显示器、音频播放装置或其他装置等。装置1300、1500还可以包括各种传感器，比如，加速度计、摄像头、光敏传感器、指纹传感器等，此 处不作限定。
可以理解，装置1300可以相当于图2对应实施例中定位设备，比如车辆160/终端180/定位服务器190，或图3中车辆100；装置1500可以相当于图2对应实施例中定位设备，比如车辆160/终端180/定位服务器190，或者装置1500可以相当于图2中几何特征地图生成设备140，或图3中车辆100。本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的系统、装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。
所述功能如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(Read-Only Memory，ROM)、随机存取存储器(Random Access Memory，RAM)、磁碟或者光盘等各种可以存储程序代码的介质。
以上实施例中所使用的术语只是为了描述特定实施例的目的，而并非旨在作为对本申请的限制。如在本申请的说明书和所附权利要求书中所使用的那样，单数表达形式“一个”、“一种”、“所述”、“上述”、“该”和“这一”旨在也包括例如“一个或多个”这种表达形式，除非其上下文中明确地有相反指示。还应当理解，在本申请以下各实施例中，“至少一个”、“一个或多个”是指一个、两个或两个以上。术语“和/或”，用于描述关联对象的关联关系，表示可以存在三种关系；例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B的情况，其中A、B可以是单数或者复数。字符“/”一般表示前后关联对象是一种“或”的关系。
在本说明书中描述的参考“一个实施例”或“一些实施例”等意味着在本申请的一个或多个实施例中包括结合该实施例描述的特定特征、结构或特点。由此，在本说明书中的不同之处出现的语句“在一个实施例中”、“在一些实施例中”、“在其他一些实施例中”、“在另外一些实施例中”等不是必然都参考相同的实施例，而是意味着“一个或多个但不是所有的实施例”，除非是以其他方式另外特别强调。术语“包括”、“包含”、“具有”及它们的变形都意味着“包括但不限于”，除非是以其他方式另外特别强调。
以上所述，仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以所述权利要求的保护范围为准。

Claims (35)

  1. 一种定位方法，其特征在于，包括：
    定位设备获取车辆通过点云采集装置采集的第一点云数据；
    所述定位设备从所述第一点云数据中提取N个第一几何特征，N为正整数；
    所述定位设备根据所述N个第一几何特征对所述车辆的第一位姿进行调整以得到所述车辆的第二位姿，所述第二位姿的精确度高于所述第一位姿的精确度。
  2. 如权利要求1所述的方法，其特征在于，所述定位设备根据所述N个第一几何特征对所述车辆的第一位姿进行调整以得到所述车辆的第二位姿，包括：
    所述定位设备根据所述N个第一几何特征和几何特征地图中的N个第二几何特征对所述车辆的第一位姿进行调整以得到所述第二位姿，所述几何特征地图为从点云地图的第二点云数据中提取的几何特征所形成的地图，所述N个第二几何特征为与所述N个第一几何特征匹配的几何特征。
  3. 如权利要求2所述的方法，其特征在于，所述定位设备根据所述第一几何特征和几何特征地图中的第二几何特征对所述车辆的第一位姿进行调整以得到所述第二位姿，包括：
    所述定位设备根据所述N个第一几何特征和所述几何特征地图中的N个第二几何特征确定几何特征之间的变换关系；
    所述定位设备根据所述几何特征之间的变换关系对所述车辆的第一位姿进行调整以得到所述第二位姿。
  4. 如权利要求3所述的方法，其特征在于，所述定位设备根据所述N个第一几何特征和所述几何特征地图中的N个第二几何特征确定几何特征之间的变换关系，包括：
    所述定位设备通过第一变换量对所述N个第一几何特征进行变换以得到N个第三几何特征，所述第三几何特征与所述第一几何特征一一对应；
    所述定位设备根据所述N个第三几何特征与所述N个第二几何特征之间的第一误差调节所述第一变换量；
    在所述第一变换量的迭代次数满足停止迭代条件或所述第一误差满足停止迭代条件时，所述定位设备得到第一目标变换量，所述第一目标变换量为满足所述停止迭代条件时的第一变换量，所述第一目标变换量用于指示所述N个第一几何特征和所述N个第二几何特征之间的变换关系。
  5. 如权利要求3所述的方法，其特征在于，所述定位设备根据所述N个第一几何特征和所述几何特征地图中的N个第二几何特征确定几何特征之间的变换关系，具体包括：
    所述定位设备通过第二变换量对所述N个第二几何特征进行变换以得到N个第四几何特征，所述第四几何特征与所述第二几何特征一一对应；
    所述定位设备所述N个第四几何特征与所述N个第一几何特征之间的第二误差调节所 述第二变换量；
    在所述第二变换量的迭代次数停止迭代条件或所述第二误差满足停止迭代条件时，所述定位设备得到第二目标变换量，所述第二目标变换量为满足所述停止迭代条件时第二变换量的逆矩阵，所述第二目标变换量用于指示所述N个第一几何特征和所述N个第二几何特征之间的变换关系。
  6. 如权利要求2所述的方法，其特征在于，所述定位设备根据所述N个第一几何特征和几何特征地图中的N个第二几何特征对所述车辆的第一位姿进行调整以得到所述第二位姿，包括：
    所述定位设备根据所述第一位姿，对所述车辆的位姿进行估计以得到多组估计位姿；
    所述定位设备根据所述N个第一几何特征和所述几何特征地图中的N个第二几何特征确定所述多组估计位姿的评分；
    所述定位设备根据所述多组估计位姿中每一组估计位姿的评分确定所述车辆的第二位姿，其中，所述每一组估计位姿的评分用于指示所述每一组估计位姿与所述第二位姿的接近程度。
  7. 如权利要求6所述的方法，其特征在于，所述定位设备根据所述N个第一几何特征和所述几何特征地图中的N个第二几何特征确定所述多组估计位姿的评分，包括：
    所述定位设备根据所述每一组估计位姿和所述N个第二几何特征，确定所述每一组估计位姿对应的所述第一参数的估计值；
    所述定位设备根据所述第一位姿和所述N个第一几何特征，确定所述第一参数的观测值；
    所述定位设备根据所述每一组估计位姿对应的第一参数的估计值与所述第一参数的观测值之间的误差确定所述每一组估计位姿的评分。
  8. 如权利要求7所述的方法，其特征在于，所述第一参数为距离、方位角和高度角中的至少一种；所述每一组估计位姿对应的第一参数的估计值为所述N个第二几何特征分别相对于所述每一组估计位姿下的车辆的第一参数；所述第一参数的观测值为所述N个第一几何特征分别相对于在所述第一位姿下的车辆的第一参数。
  9. 如权利要求6所述的方法，其特征在于，所述定位设备根据所述N个第一几何特征和所述几何特征地图中的N个第二几何特征确定所述多组估计位姿的评分，包括：
    所述定位设备通过所述每一组估计位姿与所述第一位姿之间的变换关系对所述N个第二几何特征分别进行变换得到的所述每一组估计位姿对应的N个第五几何特征，所述第二几何特征与所述第五几何特征一一对应；
    所述定位设备根据所述每一组估计位姿对应的N个第五几何特征与所述N个第一几何特征之间的误差确定所述每一组估计位姿的评分。
  10. 如权利要求6所述的方法，其特征在于，所述定位设备根据所述N个第一几何特征和所述几何特征地图中的N个第二几何特征确定所述多组估计位姿的评分，包括：
    所述定位设备通过所述每一组估计位姿与所述第一位姿之间的变换关系对所述N个第一几何特征分别进行变换得到的所述每一组估计位姿对应的N个第六几何特征，所述第一几何特征与所述第六几何特征一一对应；
    所述定位设备根据所述每一组估计位姿对应的N个第六几何特征与所述N个第二几何特征之间的误差确定所述每一组估计位姿的评分。
  11. 如权利要求1-10任一项所述的方法，其特征在于，在所述定位设备根据所述N个第一几何特征对所述车辆的第一位姿进行调整以得到所述车辆的第二位姿之前，所述方法还包括：
    所述定位设备获取所述车辆的第一位姿。
  12. 如权利要求11所述的方法，其特征在于，所述定位设备获取车辆的第一位姿，具体包括：
    所述定位设备根据上一时刻的第二位姿确定所述车辆在所述当前时刻的预测位姿，所述当前时刻的预测位姿为所述定位设备获取的车辆的第一位姿，所述上一时刻为所述当前时刻之前的时刻；或，
    所述定位设备根据定位系统确定所述车辆的第一位置以及根据惯性传感器确定所述车辆的第一姿态，所述第一位姿包括所述第一位置和所述第一姿态。
  13. 如权利要求2-12任一项所述的方法，其特征在于，所述方法还包括：所述定位设备在所述几何特征地图中查找与所述N个第一几何特征相匹配的N个第二几何特征。
  14. 如权利要求13所述的方法，其特征在于，所述定位设备在所述几何特征地图中查找与所述N个第一几何特征相匹配的N个第二几何特征，具体包括：
    所述定位设备从所述几何特征地图的第一区域中查找与所述N个第一几何特征相匹配的N个第二几何特征，所述第一区域为基于所述第一位姿确定的区域，所述第一区域不小于所述车辆的点云采集装置的扫描范围。
  15. 如权利要求1-14任一项所述的方法，其特征在于，所述第一点云数据为在由所述第一位姿确定的空间中表示的所述车辆观测到的对象的表面上的点的信息，所述N个第一几何特征中每个第一几何特征用于指示在由所述第一位姿确定的空间中所述车辆观测到的一个对象的几何特征。
  16. 如权利要求1-15任一项所述的方法，其特征在于，所述定位设备从所述第一点云数据中提取N个第一几何特征，包括：
    所述定位设备识别所述第一点云数据中的N个对象；
    所述定位设备基于所述N个对象中每个对象的点云数据确定所述每个对象的第一几何特征。
  17. 如权利要求16所述的方法，其特征在于，第一对象为所述N个对象中的任意一个对象，所述定位设备基于所述N个对象中每个对象的点云数据确定所述每个对象的第一几何特征，包括如下步骤中的一种：
    若所述第一对象的几何形状为直线，所述定位设备对所述第一对象的点云数据进行直线拟合以得到所述第一对象的第一几何特征，所述第一对象的第一几何特征为拟合得到的直线的几何特征；
    若所述第一对象的几何形状为曲线，所述定位设备对所述第一对象的点云数据进行曲线拟合以得到所述第一对象的第一几何特征，所述第一对象的第一几何特征为拟合得到的曲线的几何特征；
    若所述第一对象的几何形状为平面，所述定位设备对所述第一对象的点云数据进行平面拟合以得到所述第一对象的第一几何特征，所述第一对象的第一几何特征为拟合得到的平面的几何特征；
    若所述第一对象的几何形状为曲面，所述定位设备对所述第一对象的点云数据进行曲面拟合以得到所述第一对象的第一几何特征，所述第一对象的第一几何特征为曲面得到的曲线的几何特征。
  18. 一种定位装置，其特征在于，包括：
    第一获取单元，用于车辆通过点云采集装置采集的第一点云数据；
    特征提取单元，用于从所述第一点云数据中提取N个第一几何特征，N为正整数；
    调整单元，用于根据所述N个第一几何特征对所述车辆的第一位姿进行调整以得到所述车辆的第二位姿，所述第二位姿的精确度高于所述第一位姿的精确度。
  19. 如权利要求18所述的装置，其特征在于，所述调整单元，具体用于：
    根据所述N个第一几何特征和几何特征地图中的N个第二几何特征对所述车辆的第一位姿进行调整以得到所述第二位姿，所述几何特征地图为从点云地图的第二点云数据中提取的几何特征所形成的地图，所述N个第二几何特征为与所述N个第一几何特征匹配的几何特征。
  20. 如权利要求19所述的装置，其特征在于，所述调整单元用于根据所述第一几何特征和几何特征地图中的第二几何特征对所述车辆的第一位姿进行调整以得到所述第二位姿，包括：
    根据所述N个第一几何特征和所述几何特征地图中的N个第二几何特征确定几何特征之间的变换关系；
    根据所述几何特征之间的变换关系对所述车辆的第一位姿进行调整以得到所述第二位姿。
  21. 如权利要求20所述的装置，其特征在于，所述调整单元用于根据所述N个第一几何特征和所述几何特征地图中的N个第二几何特征确定几何特征之间的变换关系，包括：
    通过第一变换量对所述N个第一几何特征进行变换以得到N个第三几何特征，所述第三几何特征与所述第一几何特征一一对应；
    根据所述N个第三几何特征与所述N个第二几何特征之间的第一误差调节所述第一变换量；
    在所述第一变换量的迭代次数满足停止迭代条件或所述第一误差满足停止迭代条件时，得到第一目标变换量，所述第一目标变换量为满足所述停止迭代条件时的第一变换量，所述第一目标变换量用于指示所述N个第一几何特征和所述N个第二几何特征之间的变换关系。
  22. 如权利要求20所述的装置，其特征在于，所述调整单元用于根据所述N个第一几何特征和所述几何特征地图中的N个第二几何特征确定几何特征之间的变换关系，包括：
    通过第二变换量对所述N个第二几何特征进行变换以得到N个第四几何特征，所述第四几何特征与所述第二几何特征一一对应；
    根据所述第二几何特征对应的所述N个第四几何特征与所述第二几何特征对应的N个第一几何特征之间的第二误差调节所述第二变换量；
    在所述第二变换量的迭代次数停止迭代条件或所述第二误差满足停止迭代条件时，得到第二目标变换量，所述第二目标变换量为满足所述停止迭代条件时第二变换量的逆矩阵，所述第二目标变换量用于指示所述N个第一几何特征和所述N个第二几何特征之间的变换关系。
  23. 如权利要求19所述的装置，其特征在于，所述调整单元具体用于：
    根据所述第一位姿，对所述车辆的位姿进行估计以得到多组估计位姿；
    根据所述N个第一几何特征和所述几何特征地图中的N个第二几何特征确定所述多组估计位姿的评分；
    根据所述多组估计位姿中每一组估计位姿的评分确定所述车辆的第二位姿，其中，所述每一组估计位姿的评分用于指示所述每一组估计位姿与所述第二位姿的接近程度。
  24. 如权利要求23所述的装置，其特征在于，所述调整单元用于根据所述N个第一几何特征和所述几何特征地图中的N个第二几何特征确定所述多组估计位姿的评分，包括：
    根据所述每一组估计位姿和所述N个第二几何特征，确定所述每一组估计位姿对应的所述第一参数的估计值；
    根据所述第一位姿和所述N个第一几何特征，确定所述第一参数的观测值；
    根据所述每一组估计位姿对应的第一参数的估计值与所述第一参数的观测值之间的误差确定所述每一组估计位姿的评分。
  25. 如权利要求24所述的装置，其特征在于，所述第一参数为距离、方位角和高度角中的至少一种；所述每一组估计位姿对应的第一参数的估计值为所述N个第二几何特征分别相对于所述每一组估计位姿下的车辆的第一参数；所述第一参数的观测值为所述N个第一几何特征分别相对于在所述第一位姿下的车辆的第一参数。
  26. 如权利要求23所述的装置，其特征在于，所述调整单元用于根据所述N个第一几何特征和所述几何特征地图中的N个第二几何特征确定所述多组估计位姿的评分，包括：
    通过所述每一组估计位姿与所述第一位姿之间的变换关系对所述N个第二几何特征分别进行变换得到的所述每一组估计位姿对应的N个第五几何特征，所述第二几何特征与所述第五几何特征一一对应；
    根据所述每一组估计位姿对应的N个第五几何特征与所述N个第一几何特征之间的误差确定所述每一组估计位姿的评分。
  27. 如权利要求23所述的装置，其特征在于，所述调整单元用于根据所述N个第一几何特征和所述几何特征地图中的N个第二几何特征确定所述多组估计位姿的评分，包括：
    通过所述每一组估计位姿与所述第一位姿之间的变换关系对所述N个第一几何特征分别进行变换得到的所述每一组估计位姿对应的N个第六几何特征，所述第一几何特征与所述第六几何特征一一对应；
    根据所述每一组估计位姿对应的N个第六几何特征与所述N个第二几何特征之间的误差确定所述每一组估计位姿的评分。
  28. 如权利要求18-27任一项所述的装置，其特征在于，所述装置还包括：
    第二获取单元，用于在所述调整单元根据所述N个第一几何特征对所述车辆的第一位姿进行调整以得到所述车辆的第二位姿之前，获取所述车辆的第一位姿。
  29. 如权利要求28所述的装置，其特征在于，所述第二获取单元具体用于：
    根据上一时刻的第二位姿确定所述车辆在所述当前时刻的预测位姿，所述当前时刻的预测位姿为所述定位设备获取的车辆的第一位姿，所述上一时刻为所述当前时刻之前的时刻；或，
    根据定位系统确定所述车辆的第一位置以及根据惯性传感器确定所述车辆的第一姿态，所述第一位姿包括所述第一位置和所述第一姿态。
  30. 如权利要求19-29任一项所述的装置，其特征在于，所述装置还包括：
    匹配单元，用于在所述几何特征地图中查找与所述N个第一几何特征相匹配的N个第二几何特征。
  31. 如权利要求30所述的装置，其特征在于，所述匹配单元具体用于：
    从所述几何特征地图的第一区域中查找与所述N个第一几何特征相匹配的N个第二几 何特征，所述第一区域为基于所述第一位姿确定的区域，所述第一区域不小于所述车辆的点云采集装置的扫描范围。
  32. 如权利要求18-31任一项所述的装置，其特征在于，所述第一获取单元具体用于：
    识别所述第一点云数据中的N个对象；
    基于所述N个对象中每个对象的点云数据确定所述每个对象的第一几何特征。
  33. 一种定位装置，其特征在于，包括：处理器和存储器，所述存储器用于存储程序，所述处理器执行所述存储器存储的程序，当所述存储器存储的程序被执行时，可实现如权利要求1-17任一项所述的方法。
  34. 一种车辆，其特征在于，包括：点云采集装置、处理器和存储器，所述处理器通过总线连接到所述点云采集装置，所述点云采集装置用于采集点云数据，所述存储器用于存储程序，所述处理器执行所述存储器存储的程序，当所述存储器存储的程序被执行时，可实现如权利要求1-17任一项所述的方法。
  35. 一种计算机可读存储介质，其特征在于，所述计算机存储介质存储有计算机程序，所述计算机程序包括程序指令，所述程序指令当被处理器执行时，可实现如权利要求1-17任一项所述的方法。
PCT/CN2019/105810 2019-09-12 2019-09-12 定位方法、装置及系统 Ceased WO2021046829A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
PCT/CN2019/105810 WO2021046829A1 (zh) 2019-09-12 2019-09-12 定位方法、装置及系统
EP19930161.5A EP3819673A4 (en) 2019-09-12 2019-09-12 POSITIONING METHOD, DEVICE AND SYSTEM
CN201980055540.0A CN112639882B (zh) 2019-09-12 2019-09-12 定位方法、装置及系统
US17/689,642 US20220198706A1 (en) 2019-09-12 2022-03-08 Positioning method, apparatus, and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2019/105810 WO2021046829A1 (zh) 2019-09-12 2019-09-12 定位方法、装置及系统

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/689,642 Continuation US20220198706A1 (en) 2019-09-12 2022-03-08 Positioning method, apparatus, and system

Publications (1)

Publication Number Publication Date
WO2021046829A1 true WO2021046829A1 (zh) 2021-03-18

Family

ID=74866963

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/105810 Ceased WO2021046829A1 (zh) 2019-09-12 2019-09-12 定位方法、装置及系统

Country Status (4)

Country Link
US (1) US20220198706A1 (zh)
EP (1) EP3819673A4 (zh)
CN (1) CN112639882B (zh)
WO (1) WO2021046829A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115420290A (zh) * 2022-08-16 2022-12-02 清华大学 非合作目标机动检测方法、装置、设备及计算机存储介质
CN116229040A (zh) * 2022-07-15 2023-06-06 深圳市速腾聚创科技有限公司 目标区域的定位方法和目标区域的定位装置
CN116958568A (zh) * 2023-07-06 2023-10-27 中国测绘科学研究院 一种附加方位约束的点云线性特征稳健提取方法

Families Citing this family (33)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018031678A1 (en) * 2016-08-09 2018-02-15 Nauto Global Limited System and method for precision localization and mapping
US10816354B2 (en) 2017-08-22 2020-10-27 Tusimple, Inc. Verification module system and method for motion-based lane detection with multiple sensors
US12270661B2 (en) 2018-02-14 2025-04-08 Tusimple, Inc. Lane marking localization and fusion
US11009356B2 (en) 2018-02-14 2021-05-18 Tusimple, Inc. Lane marking localization and fusion
US11009365B2 (en) 2018-02-14 2021-05-18 Tusimple, Inc. Lane marking localization
EP4020428A4 (en) * 2019-08-28 2022-10-12 Huawei Technologies Co., Ltd. METHOD AND DEVICE FOR TRACK RECOGNITION AND COMPUTING DEVICE
EP3893150A1 (en) * 2020-04-09 2021-10-13 Tusimple, Inc. Camera pose estimation techniques
CN114063090B (zh) * 2020-07-29 2025-05-09 北京图森未来科技有限公司 一种可移动设备的定位方法、装置及可移动设备
EP4016456B1 (en) * 2020-12-18 2025-10-15 Augmented Robotics GmbH Method and system for determining a three dimensional position
CN113935425B (zh) * 2021-10-21 2024-08-16 中国船舶集团有限公司第七一一研究所 物体识别方法、装置、终端和存储介质
CN116266356A (zh) * 2021-12-17 2023-06-20 影石创新科技股份有限公司 全景视频转场渲染方法、装置和计算机设备
DE102022203267A1 (de) 2022-04-01 2023-10-05 Robert Bosch Gesellschaft mit beschränkter Haftung Verfahren zur Lokalisierung eines Fahrzeugs mittels entfernungsbasierter Sensordaten des Fahrzeugs
CN114794992B (zh) * 2022-06-07 2024-01-09 深圳甲壳虫智能有限公司 充电座、机器人的回充方法和扫地机器人
US12241757B1 (en) * 2022-06-24 2025-03-04 Aurora Operations, Inc Generation of weighted map data for autonomous vehicle localization
EP4312054A1 (en) * 2022-07-29 2024-01-31 GM Cruise Holdings LLC Radar point cloud multipath reflection compensation
CN115164855B (zh) * 2022-08-05 2024-09-13 珠海格力智能装备有限公司 一种托盘识别方法、装置、设备及可读存储介质
CN115859765B (zh) * 2022-09-29 2023-12-08 中山大学 城市扩张的预测方法、装置、设备及存储介质
CN115862321A (zh) * 2022-11-21 2023-03-28 东软睿驰汽车技术（沈阳）有限公司 场景数据获取方法、装置、设备及存储介质
US12288358B2 (en) 2022-12-02 2025-04-29 Automotive Research & Testing Center Method of sensor fusion for harmonizing data from multiple data sources
CN116182877A (zh) * 2023-02-15 2023-05-30 酷哇科技有限公司 基于路牙特征提取和匹配的车辆定位方法及系统
CN116580088A (zh) * 2023-04-10 2023-08-11 深圳市正浩创新科技股份有限公司 回充方法、回充装置、自移动设备以及存储介质
CN116823944A (zh) * 2023-05-24 2023-09-29 中山大学 一种融合路侧稀疏点云和道路特征的车辆位姿估计方法
CN116748698B (zh) * 2023-06-09 2026-03-03 湖北亿纬动力有限公司 极片模切方法
CN116977416B (zh) * 2023-06-13 2026-02-03 上海富欣智能交通控制有限公司 轨道交通的车辆位姿估计方法和系统
US12586308B2 (en) * 2023-07-13 2026-03-24 Nvidia Corporation Generating object representations using neural networks for autonomous systems and applications
US12480781B2 (en) * 2023-07-26 2025-11-25 GM Global Technology Operations LLC Crowd-sourcing lane line maps for a vehicle
CN116993828B (zh) * 2023-08-02 2025-08-08 北京工业大学 基于激光雷达点云聚类的点云匹配定位方法、系统及产品
US12566069B2 (en) * 2023-08-11 2026-03-03 Torc Robotics, Inc. Localizing vehicles using retroreflective surfaces
CN117421434B (zh) * 2023-09-08 2025-06-03 广州汽车集团股份有限公司 点云定位地图的存储方法、装置、电子设备及存储介质
US12449271B2 (en) 2023-12-06 2025-10-21 GM Global Technology Operations LLC Crowd-sourcing lane line maps for a vehicle
CN118628294B (zh) * 2024-08-14 2025-01-07 广东省建筑设计研究院集团股份有限公司 一种基于既有建筑结构的建筑巡检方法及系统
CN119270298B (zh) * 2024-12-11 2025-02-28 江苏智能无人装备产业创新中心有限公司 自动驾驶卡车的铰接角实时测算方法、系统、设备及介质
CN120411158B (zh) * 2025-07-03 2025-09-23 武汉大学 针对弱结构长隧道环境的激光惯性里程计算方法及系统

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103900583A (zh) * 2012-12-25 2014-07-02 联想(北京)有限公司 用于即时定位与地图构建的设备和方法
CN107430183A (zh) * 2014-12-26 2017-12-01 赫尔环球有限公司 用于装置的定位的几何指纹识别
CN108732603A (zh) * 2017-04-17 2018-11-02 百度在线网络技术（北京）有限公司 用于定位车辆的方法和装置
CN109059906A (zh) * 2018-06-26 2018-12-21 上海西井信息科技有限公司 车辆定位方法、装置、电子设备、存储介质
US20190079193A1 (en) * 2017-09-13 2019-03-14 Velodyne Lidar, Inc. Multiple Resolution, Simultaneous Localization and Mapping Based On 3-D LIDAR Measurements
CN109887028A (zh) * 2019-01-09 2019-06-14 天津大学 一种基于点云数据配准的无人车辅助定位方法
CN109901202A (zh) * 2019-03-18 2019-06-18 成都希德瑞光科技有限公司 一种基于点云数据的机载系统位置修正方法
CN110095752A (zh) * 2019-05-07 2019-08-06 百度在线网络技术（北京）有限公司 定位方法、装置、设备和介质

Family Cites Families (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5538667B2 (ja) * 2007-04-26 2014-07-02 キヤノン株式会社 位置姿勢計測装置及びその制御方法
NO327279B1 (no) * 2007-05-22 2009-06-02 Metaio Gmbh Kamerapositurestimeringsanordning og- fremgangsmate for foroket virkelighetsavbildning
JP5430456B2 (ja) * 2010-03-16 2014-02-26 キヤノン株式会社 幾何特徴抽出装置、幾何特徴抽出方法、及びプログラム、三次元計測装置、物体認識装置
JP6101134B2 (ja) * 2013-04-01 2017-03-22 キヤノン株式会社 情報処理装置及び情報処理方法
CN103268729B (zh) * 2013-05-22 2015-08-19 北京工业大学 基于混合特征的移动机器人级联式地图创建方法
US9803985B2 (en) * 2014-12-26 2017-10-31 Here Global B.V. Selecting feature geometries for localization of a device
US10989542B2 (en) * 2016-03-11 2021-04-27 Kaarta, Inc. Aligning measured signal data with slam localization data and uses thereof
US10670725B2 (en) * 2017-07-25 2020-06-02 Waymo Llc Determining yaw error from map data, lasers, and cameras
CN109425365B (zh) * 2017-08-23 2022-03-11 腾讯科技（深圳）有限公司 激光扫描设备标定的方法、装置、设备及存储介质
JP7278263B2 (ja) * 2017-09-04 2023-05-19 コモンウェルス サイエンティフィック アンド インダストリアル リサーチ オーガナイゼーション 位置特定を実施する際に使用するための方法およびシステム
US10438371B2 (en) * 2017-09-22 2019-10-08 Zoox, Inc. Three-dimensional bounding box from two-dimensional image and point cloud data
CN107917712B (zh) * 2017-11-16 2020-07-28 苏州艾吉威机器人有限公司 一种同步定位与地图构建方法及设备
CN108345005B (zh) * 2018-02-22 2020-02-07 重庆大学 巷道掘进机的实时连续自主定位定向系统及导航定位方法
CN109297510B (zh) * 2018-09-27 2021-01-01 百度在线网络技术（北京）有限公司 相对位姿标定方法、装置、设备及介质
CN109932713B (zh) * 2019-03-04 2021-07-09 北京旷视科技有限公司 定位方法、装置、计算机设备、可读存储介质和机器人

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103900583A (zh) * 2012-12-25 2014-07-02 联想(北京)有限公司 用于即时定位与地图构建的设备和方法
CN107430183A (zh) * 2014-12-26 2017-12-01 赫尔环球有限公司 用于装置的定位的几何指纹识别
CN108732603A (zh) * 2017-04-17 2018-11-02 百度在线网络技术（北京）有限公司 用于定位车辆的方法和装置
US20190079193A1 (en) * 2017-09-13 2019-03-14 Velodyne Lidar, Inc. Multiple Resolution, Simultaneous Localization and Mapping Based On 3-D LIDAR Measurements
CN109059906A (zh) * 2018-06-26 2018-12-21 上海西井信息科技有限公司 车辆定位方法、装置、电子设备、存储介质
CN109887028A (zh) * 2019-01-09 2019-06-14 天津大学 一种基于点云数据配准的无人车辅助定位方法
CN109901202A (zh) * 2019-03-18 2019-06-18 成都希德瑞光科技有限公司 一种基于点云数据的机载系统位置修正方法
CN110095752A (zh) * 2019-05-07 2019-08-06 百度在线网络技术（北京）有限公司 定位方法、装置、设备和介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3819673A4 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116229040A (zh) * 2022-07-15 2023-06-06 深圳市速腾聚创科技有限公司 目标区域的定位方法和目标区域的定位装置
CN116229040B (zh) * 2022-07-15 2025-10-28 深圳市速腾聚创科技有限公司 目标区域的定位方法和目标区域的定位装置
CN115420290A (zh) * 2022-08-16 2022-12-02 清华大学 非合作目标机动检测方法、装置、设备及计算机存储介质
CN115420290B (zh) * 2022-08-16 2024-05-31 清华大学 非合作目标机动检测方法、装置、设备及计算机存储介质
CN116958568A (zh) * 2023-07-06 2023-10-27 中国测绘科学研究院 一种附加方位约束的点云线性特征稳健提取方法

Also Published As

Publication number Publication date
CN112639882A (zh) 2021-04-09
CN112639882B (zh) 2021-12-14
US20220198706A1 (en) 2022-06-23
EP3819673A4 (en) 2021-07-21
EP3819673A1 (en) 2021-05-12

Similar Documents

Publication Publication Date Title
CN112639882B (zh) 定位方法、装置及系统
US12223740B2 (en) Image processing method, network training method, and related device
CN113168708B (zh) 车道线跟踪方法和装置
CN113792566B (zh) 一种激光点云的处理方法及相关设备
CN112740268B (zh) 目标检测方法和装置
US12367686B2 (en) Road structure detection method and apparatus
CN114255275B (zh) 一种构建地图的方法及计算设备
JP2019527832A (ja) 正確な位置特定およびマッピングのためのシステムおよび方法
WO2022033089A1 (zh) 确定检测对象的三维信息的方法及装置
EP4206731A1 (en) Target tracking method and device
US20230342883A1 (en) Image processing method and apparatus, and storage medium
EP4307251B1 (en) Mapping method, vehicle, computer readable storage medium, and chip
WO2021159397A1 (zh) 车辆可行驶区域的检测方法以及检测装置
US20250292550A1 (en) Data Processing Method and Apparatus
WO2022142839A1 (zh) 一种图像处理方法、装置以及智能汽车
CN116106931A (zh) 一种涉水预警的方法及相关装置
CN114842075B (zh) 数据标注方法、装置、存储介质及车辆
US20250391273A1 (en) Data Processing Method, Readable Storage Medium, and Electronic Device
CN113066124A (zh) 一种神经网络的训练方法以及相关设备
CN119987341B (zh) 一种路径优化方法及相关装置
WO2025180245A9 (zh) 一种车辆的定位方法、神经网络的训练方法以及相关设备
WO2026040317A1 (zh) 一种地图信息的获取方法以及相关装置
CN120765797A (zh) 一种制图方法以及相关设备
CN121026164A (zh) 一种地图信息的处理方法、地图信息的获取方法以及设备
CN116452762A (zh) 构建地图模型的方法、定位方法及装置

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2019930161

Country of ref document: EP

Effective date: 20201130

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19930161

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE