WO2023207123A1 - 一种点云数据的分层方法、装置、设备、介质及车辆 - Google Patents

一种点云数据的分层方法、装置、设备、介质及车辆 Download PDF

Info

Publication number
WO2023207123A1
WO2023207123A1 PCT/CN2022/137936 CN2022137936W WO2023207123A1 WO 2023207123 A1 WO2023207123 A1 WO 2023207123A1 CN 2022137936 W CN2022137936 W CN 2022137936W WO 2023207123 A1 WO2023207123 A1 WO 2023207123A1
Authority
WO
WIPO (PCT)
Prior art keywords
point cloud
index value
cloud data
value range
vehicle
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/CN2022/137936
Other languages
English (en)
French (fr)
Inventor
李江龙
罗金辉
单乐
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.)
Beijing Chusudu Technology Co Ltd
Original Assignee
Beijing Chusudu Technology 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 Beijing Chusudu Technology Co Ltd filed Critical Beijing Chusudu Technology Co Ltd
Priority to EP22908817.4A priority Critical patent/EP4290467A4/en
Priority to US18/340,990 priority patent/US20230331238A1/en
Publication of WO2023207123A1 publication Critical patent/WO2023207123A1/zh
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
    • G06T17/00—Three-dimensional [3D] modelling for computer graphics

Definitions

  • Embodiments of the present invention relate to the field of autonomous driving technology, and specifically, to a layered method, device, equipment, medium and vehicle for point cloud data.
  • Point cloud data of different layer heights can be used to create corresponding high-precision maps based on the point cloud data of each layer height.
  • the main method is to perform plane fitting and segmentation on the vehicle's driving trajectory, and find each plane as a layer.
  • the same layer is often not on the same plane. Therefore, the above solution is prone to layering errors, which affects the subsequent production of high-precision maps.
  • Embodiments of the present invention provide a point cloud data layering method, device, equipment, medium and vehicle to overcome the problem of inaccurate layering of point cloud data.
  • embodiments of the present invention provide a hierarchical method for point cloud data, including:
  • the trajectory to be processed is divided into ramp areas and non-ramp areas;
  • the embodiment of the present invention takes into account the situation when the vehicle is driving on a slope road or a non-slope road.
  • the difference in vehicle attitude is determined by determining the pitch angle used to characterize the vehicle attitude, and based on the difference in pitch angle when the vehicle is driving on a non-slope road and when driving on a slope road, the vehicle driving trajectory is divided into a ramp area and a non-slope area.
  • the method provided by the embodiment of the present invention further includes:
  • the point cloud data in different layers are visually displayed, and planes corresponding to different layer heights are generated for creating maps corresponding to each plane.
  • point cloud data in different layers can be visually displayed, so that different workers can construct maps in different layers to improve mapping efficiency.
  • the created high-precision map can be used in the positioning process of autonomous vehicles. In the scenarios of viaducts, parking lots, especially underground parking lots, it can provide autonomous vehicles with information about the number of layers of the road they are currently on. .
  • the trajectory to be processed is divided into a ramp area and a non-ramp area, including:
  • the point cloud data corresponding to the pitch angle of the vehicle that is greater than the preset angle is used as the first point cloud data of the target vehicle when it is going uphill or downhill;
  • the trajectories to be processed corresponding to the set of all first point cloud data are regarded as ramp areas, and the trajectories to be processed corresponding to the set of all second point cloud data except the first point cloud data are regarded as non-ramp areas. .
  • the trajectory to be processed corresponding to the set of all first point cloud data is used as the ramp area
  • the trajectory to be processed corresponding to the set of all second point cloud data except the first point cloud data is used as Non-ramp areas
  • the point cloud frame number index values corresponding to all first point cloud data are divided into multiple first index value ranges, where the number of first index value ranges is used Indicates the number of times the target vehicle has gone uphill or downhill;
  • the first index value range is expanded by adding several index values at both ends of the index value range to obtain an expanded target index value range, and each target index value range is obtained.
  • the trajectories to be processed corresponding to the index value range are respectively used as ramp areas;
  • the point cloud frame number index values corresponding to the second point cloud data except the first point cloud data are Divide the method into multiple second index value ranges, and use the trajectories to be processed corresponding to each second index value range as non-ramp areas.
  • the embodiment of the present invention combines each frame of point cloud data according to the relationship between the vehicle pitch angle and the preset angle.
  • the frame point cloud data is initially divided into the first point cloud data corresponding to the ramp area and the second point cloud data corresponding to the non-ramp area.
  • the ramp area may be represented by a first index value range corresponding to the first point cloud data.
  • the embodiment of the present invention expands each first index value range corresponding to the first point cloud data, so that the points corresponding to the ramp area can be The determination of cloud data is more accurate.
  • the first index value range is expanded by adding several index values at both ends of the index value range to obtain an expanded target index value range, including:
  • the monotonicity between the target vehicle height information corresponding to each index value in the first index value range add index values that comply with the monotonicity at both ends of the index value range to adjust the first index value.
  • the range is expanded to obtain an expanded target index value range, wherein the monotonicity includes monotonic increase or monotonic decrease.
  • the monotonicity between the target vehicle height information corresponding to each index value in the first index value range add index values that comply with the monotonicity at both ends of the index value range, so as to The first index value range is expanded to obtain the expanded target index value range, including:
  • the index value to be increased is combined with the index value in the original first index value range to form an expanded target index value range.
  • the embodiment of the present invention determines the slope corresponding to the first index value range corresponding to the initially divided ramp area based on the monotonically increasing or monotonically decreasing property of the height information of the target vehicle when driving in the ramp area.
  • the range of the first index value corresponding to the ramp area can be expanded by determining other index values whose height difference from the height information of the intermediate index value conforms to the monotonically increasing property. , so that the expanded target index value range can correspond to a completed ramp area, thereby effectively improving the determination accuracy of the ramp area.
  • the vehicle pitch angle is the vehicle pitch angle after filtering.
  • the embodiment of the present invention solves the interference caused by the slight jitter of the vehicle during driving by filtering the vehicle pitch angle, and improves the accuracy of the vehicle pitch angle, thereby making the subsequent ramp area and Non-ramp areas are demarcated more accurately.
  • stratify non-ramp zoning according to height information including:
  • the non-ramp area is divided into multiple layers according to the order of the average height information of different sub-areas from large to small.
  • the embodiment of the present invention first determines the ramp area from the trajectory to be processed, and then determines the remaining areas in the trajectory to be processed. It is a non-ramp area, and the non-ramp area is divided into multiple layers in order from high to low by height clustering algorithm. Compared with the method of plane fitting and segmentation of the trajectory to be processed in the related art, since the road surface is not completely horizontal, it is easy to fit the road surface into multiple layers. In the layering process, the embodiment of the present invention does not pay attention to Whether the road surface corresponding to the non-slope area of each layer is level, and there is no need to consider whether point cloud data of other planes are used during plane fitting.
  • the point cloud layering scheme provided by the embodiment of the present invention effectively improves the accuracy of point cloud data layering. sex.
  • determine the vehicle pitch angle of the target vehicle relative to the horizontal plane when collecting each frame of point cloud data including:
  • the vehicle body coordinate system is a coordinate system that is fixedly connected to the vehicle body
  • the standard coordinate system is the coordinate system corresponding to the horizontal plane
  • the vehicle pitch angle of the target vehicle relative to the horizontal plane is determined when each frame of point cloud data is collected.
  • the multiple sensors include IMU, GPS, radar and/or image sensors.
  • embodiments of the present invention also provide a layering device for point cloud data, including:
  • the pitch angle determination module is configured to, for each frame of point cloud data corresponding to the trajectory to be processed, determine the vehicle pitch angle of the target vehicle relative to the horizontal plane when collecting each frame of point cloud data, wherein the standard coordinate system is used to represent the horizontal plane ;
  • the area division module is configured to divide the trajectory to be processed into a ramp area and a non-ramp area based on the size relationship between the vehicle pitch angles corresponding to each frame of point cloud data;
  • a layering module configured to layer non-ramp areas according to height information.
  • the device provided by the embodiment of the present invention also includes:
  • the mapping module is configured to, after stratifying the non-ramp areas according to height information, visually display the point cloud data in different layers according to the stratification results of the non-ramp areas, and generate images corresponding to different layer heights. Planes, used to create maps corresponding to each plane.
  • regional division module including:
  • the point cloud data dividing unit is configured to, in each frame of point cloud data, use the point cloud data corresponding to the pitch angle of the vehicle greater than the preset angle as the first point cloud of the target vehicle when it is going uphill or downhill. data;
  • the area dividing unit is configured to use the to-be-processed trajectories corresponding to the set of all first point cloud data as the ramp area, and to use the to-be-processed trajectories corresponding to the set of all second point cloud data except the first point cloud data. Treat tracks as non-ramp areas.
  • regional division units include:
  • the index value range determination subunit is configured to divide the point cloud frame number index values corresponding to all first point cloud data into multiple first index value ranges based on the continuity between point cloud frame number index values, where, The number of the first index value range is used to represent the number of times the target vehicle goes uphill or downhill;
  • the ramp area determination subunit is configured to expand the first index value range for any first index value range by adding several index values at both ends of the index value range to obtain the expanded range.
  • target index value range, and the trajectories to be processed corresponding to each target index value range are used as ramp areas;
  • the non-slope area division subunit is configured to divide the second point cloud data in addition to the first point cloud data based on the continuity between the multiple target index value ranges corresponding to the first point cloud data and the continuity between the point cloud frame number index values.
  • the point cloud frame number index value corresponding to the point cloud data is divided into a plurality of second index value ranges, and the trajectories to be processed corresponding to each second index value range are respectively regarded as non-slope areas.
  • ramp area determination subunits include:
  • the ramp area determination component is configured to, for any first index value range, according to the monotonicity between the target vehicle height information corresponding to each index value in the first index value range, at the two end positions of the index value range Index values that comply with the monotonicity are respectively added to expand the first index value range to obtain the expanded target index value range, and the to-be-processed trajectories corresponding to each target index value range are respectively used as ramp areas. ; Among them, monotonicity includes monotonic increase or monotonic decrease.
  • ramp area determination component specifically configured as:
  • the increased index value is combined with the index value in the original first index value range to form an expanded target index value range, and the to-be-processed trajectory corresponding to each target index value range is respectively as ramp areas.
  • the vehicle pitch angle is the vehicle pitch angle after filtering.
  • layered modules include:
  • the sub-region determination unit is configured to determine sub-regions with the same average height information from the non-slope area, wherein the point cloud frame number index value corresponding to each sub-region is continuous;
  • the layering unit is configured to divide the non-ramp area into multiple layers in descending order of the average height information of different sub-areas.
  • pitch angle determination module specifically configured:
  • the vehicle body coordinate system is a coordinate system that is fixedly connected to the vehicle body
  • the standard coordinate system is the coordinate system corresponding to the horizontal plane
  • the vehicle pitch angle of the target vehicle relative to the horizontal plane is determined when each frame of point cloud data is collected.
  • the multiple sensors include IMU, GPS, radar and/or image sensors.
  • embodiments of the present invention provide a storage medium on which a computer program is stored. When the program is executed by a processor, the method described in any embodiment of the first aspect is implemented.
  • an embodiment of the present invention provides an electronic device.
  • the electronic device includes:
  • processors one or more processors
  • a storage device for storing one or more programs
  • the one or more processors When the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the method described in any embodiment of the first aspect.
  • embodiments of the present invention provide a vehicle, which includes a device as described in any embodiment of the second aspect, or an electronic device as described in the fourth aspect.
  • embodiments of the present invention provide a computer program.
  • the computer program includes program instructions.
  • the program instructions are executed by a computer, the method as described in any embodiment of the first aspect is implemented.
  • Figure 1a is a flow chart of a point cloud data layering method provided in Embodiment 1 of the present invention.
  • Figure 1b is a schematic diagram of the relationship between the vehicle pitch angle and the point cloud frame number index value provided in Embodiment 1 of the present invention
  • Figure 1c is a side view of a basement track provided in Embodiment 1 of the present invention.
  • Figure 2 is a flow chart of a point cloud data layering method provided in Embodiment 2 of the present invention.
  • Figure 3 is a flow chart of a point cloud data layering method provided in Embodiment 3 of the present invention.
  • Figure 4 is a structural block diagram of a point cloud data layering device provided in Embodiment 4 of the present invention.
  • Figure 5 is a structural block diagram of an electronic device provided in Embodiment 5 of the present invention.
  • Figure 6 is a schematic diagram of a vehicle provided in Embodiment 5 of the present invention.
  • Embodiments of the present invention disclose a point cloud data layering method, device, equipment, medium and vehicle. Each is explained in detail below.
  • Figure 1a is a flow chart of a point cloud data layering method provided in Embodiment 1 of the present invention.
  • This method can be applied to vehicle-mounted computers, vehicle-mounted industrial personal computers (IPC) and other vehicle-mounted terminals, and can also be applied to server, this embodiment of the present invention does not limit this.
  • the method provided in this embodiment can be applied to the hierarchical processing of point cloud data in viaducts, parking lots, especially underground parking lots, and the hierarchical results can be used for map creation or vehicle positioning.
  • the method provided in this embodiment can be executed by a hierarchical device of point cloud data, and the device can be implemented by software and/or hardware. As shown in Figure 1a, the method provided by this embodiment specifically includes:
  • the trajectory to be processed is a trajectory generated by the target vehicle driving in the target scene, and the trajectory to be processed includes multiple trajectory points sorted in chronological order.
  • the point cloud data is the point cloud data corresponding to the target scene observed by the target vehicle at each trajectory point.
  • the target scenes can be viaducts, parking lots, especially underground parking lots and other scenes.
  • the target vehicle can be a map collection vehicle equipped with a variety of sensor devices.
  • point cloud data there are many ways to obtain point cloud data.
  • a ranging sensor such as a radar or laser scanner, etc.
  • An image acquisition device such as a depth camera or a binocular camera, etc. collects the image of each trajectory point, and then obtains the point cloud data of each trajectory point based on the image of each trajectory point.
  • the embodiments of the present invention do not limit the specific acquisition method of point cloud data. Any method that can obtain the point cloud data corresponding to the target scene observed at each trajectory point can be applied to the embodiments of the present invention.
  • the vehicle attitude and the horizontal plane are parallel.
  • the included angle can be represented by the pitch angle of the vehicle attitude.
  • a standard coordinate system that represents the horizontal plane can be selected first.
  • the standard coordinate system can be a coordinate system established with a point on the local horizontal plane as the origin and the horizontal plane as the xoy plane.
  • the standard coordinate system may be the ENU coordinate system (Northeast Celestial Coordinate System).
  • the initial pitch angle of the vehicle relative to the standard coordinate system can be obtained based on the relative relationship between the vehicle body coordinate system and the standard coordinate system.
  • the vehicle body coordinate system is defined on the vehicle body and is a coordinate system fixedly connected to the vehicle body. For example, it can be defined at the center of the rear axle of the vehicle body. If the vehicle is on a horizontal plane, the vehicle body coordinate system is also in a horizontal state.
  • the IMU can sense the direction of gravity
  • the IMU can be used to determine the initial pitch of the vehicle attitude of the target vehicle relative to the horizontal plane when collecting each frame of point cloud data. horn.
  • the attitude of the vehicle can be determined based on the fused data of multiple sensors, thereby determining the initial vehicle pitch angle of the attitude relative to the horizontal plane.
  • multi-sensors may include IMU, GPS (Global Positioning System), radar and/or image sensors.
  • the initial pitch angle can be filtered to solve the interference caused by the slight jitter of the vehicle during driving, thereby improving the accuracy of the vehicle pitch angle.
  • the mean filtering method can be used, specifically: for the initial vehicle pitch angle corresponding to the point cloud data of the current frame, the point cloud data of the two frames before and after this frame of point cloud data corresponds to Perform mean filtering on the vehicle pitch angle corresponding to the current frame point cloud data, and use the pitch angle value obtained after mean filtering as the vehicle pitch angle corresponding to the current frame point cloud data.
  • This mean filtering process can be performed by It is represented by the following formula:
  • p[i] (back) (p[i-2]+p[i-1]+p[i] (front) +p[i+1]+p[i+2])/5;
  • i represents the index value of the current frame point cloud data
  • p[i] (before) represents the vehicle pitch angle corresponding to the current frame point cloud data before filtering
  • p[i] (after) represents the current frame point after filtering The vehicle pitch angle corresponding to the cloud data.
  • Gaussian filtering or median filtering can also be used to filter the vehicle pitch angle corresponding to each frame of point cloud data. This embodiment does not specifically limit the filtering method.
  • FIG. 1b is a schematic diagram of the relationship between the vehicle pitch angle and the point cloud frame number index value provided in Embodiment 1 of the present invention.
  • the abscissa represents the point cloud frame number index value
  • the ordinate represents the vehicle Pitch angle.
  • the pitch angle of the four peak areas A framed in the figure has a large range of numerical changes, which corresponds to the pitch angle of the vehicle when driving on a slope section.
  • the ramp section may be a vehicle traveling from a higher-level road section to a lower-level road section, or it may be traveling from a lower-level road section to a higher-level road section.
  • the numerical variation range of the pitch angle in area B framed in Figure 1b is smaller and corresponds to the non-slope section.
  • the target vehicle is collected in each frame by calculating
  • the vehicle pitch angle relative to the horizontal plane in the point cloud data can be used to determine the driving trajectory corresponding to the off-ramp section, that is, the trajectory corresponding to the peak area A shown in Figure 1b can be found, and then the non-ramp in the trajectory to be processed can be determined
  • the trajectory corresponding to the area is the trajectory corresponding to area B as shown in Figure 1b.
  • FIG. 1c is a side view of a basement track provided in Embodiment 1 of the present invention.
  • a vehicle drives from the above-ground road into the underground B1-level garage, or drives from the underground B1-level garage onto the above-ground road, or drives through the underground B1-level garage into the underground B2-level garage, it will pass through a slope section.
  • the vehicle attitude is parallel to the horizontal plane, and the vehicle pitch angle usually changes within a small value range.
  • the vehicle pitch angle usually changes within a larger value range. Based on the above principle, it can be determined whether the vehicle is traveling in a ramp area or a non-ramp area based on the relationship between the vehicle pitch angles corresponding to each frame of point cloud data.
  • an angle can be preset as the critical angle of the vehicle in the non-ramp area and the ramp area.
  • the preset angle can be determined based on the empirical value of the pitch angle when the vehicle travels on non-slope sections and slope sections many times. For example, when the vehicle is traveling across floors, as shown in Figure 1b, the vehicle pitch angle usually changes between 5° and 10°. In this embodiment, the preset angle can be set to 5°.
  • the point cloud data corresponding to the pitch angle of the vehicle that is greater than the preset angle can be used as the first point cloud data of the target vehicle when it is going uphill or downhill. Since non-ramp areas at different heights are connected through the ramp area, and due to the continuity of the trajectory to be processed, after determining the first point cloud data corresponding to the ramp area, the other points except the first point cloud data can be The trajectories to be processed corresponding to the set of all second point cloud data are regarded as non-ramp areas.
  • the trajectory to be processed since the trajectory to be processed includes multiple trajectory points sorted in chronological order, after determining the ramp area according to the vehicle pitch angle, the trajectory to be processed can be divided into multiple sub-regions according to the chronological order. That is, it is divided into non-ramp sub-regions, ramp sub-regions, non-ramp sub-regions and ramp sub-regions, etc., where the number of point cloud frames corresponding to each sub-region is continuous.
  • the non-ramp area will contain multiple non-ramp sub-areas with the same height value.
  • the non-ramp sub-area with the average height value within the set range can be regarded as the non-ramp road area of the same layer, so that the non-ramp area can be Divided into multiple layers.
  • the average height value of the non-ramp sub-region can be determined based on the vehicle height information corresponding to each point cloud frame. Among them, vehicle height information can be obtained through GPS.
  • the non-ramp area can be divided into multiple non-ramp sub-areas according to the continuity of the number of point cloud frames in each sub-area.
  • the average height of each non-ramp sub-region By calculating the average height of each non-ramp sub-region and clustering each non-ramp sub-region according to the average height, non-ramp sub-regions with the same height information can be obtained, that is, the non-ramp sub-regions with the same height that the vehicle has traveled in different time periods can be obtained.
  • Road segments are grouped into one category. Among them, the clustering height can be set to 2 meters.
  • the non-ramp area After completing the clustering of non-ramp sub-areas of the same height, the non-ramp area can be divided into multiple layers according to the order of the average height information of different non-ramp sub-areas from large to small.
  • the layering process of this embodiment it does not pay attention to whether the road surface corresponding to the non-slope area of each layer is level, nor does it need to consider whether point cloud data of other planes are used during plane fitting.
  • the point cloud layering scheme provided in this embodiment effectively improves the point cloud data analysis. layer accuracy.
  • the pitch angle used to characterize the vehicle attitude is determined, and the vehicle is driven on a non-slope road section and on a slope road section.
  • the trajectory to be processed can be divided into ramp areas and non-ramp areas, and then the non-ramp areas can be processed in layers.
  • the technical solution provided by this embodiment avoids dividing point cloud data of different layers into the same layer of point cloud data. problem, which improves the accuracy of point cloud data layering, thereby helping to improve the accuracy of subsequent high-precision map production.
  • Figure 2 is a flow chart of a point cloud data layering method provided in Embodiment 2 of the present invention. Based on the above embodiment, this embodiment details the specific division process of ramp areas and non-ramp areas. As shown in Figure 2, the method provided by this embodiment includes:
  • step S210 For the specific implementation of step S210, reference may be made to the description of the above embodiments, which will not be described again here.
  • each frame of point cloud data use the point cloud data corresponding to the pitch angle of the vehicle greater than the preset angle as the first point cloud data of the target vehicle when it is going uphill or downhill.
  • the third point cloud data when determining the first point cloud data of the target vehicle when it is going uphill or downhill, and the second point cloud data corresponding to the non-slope area except the first point cloud data, the third point cloud data can be determined.
  • the point cloud frame number index value corresponding to the point cloud data and the point cloud frame number index value corresponding to the second point cloud data are stored.
  • the ramp sections can be determined first, and then based on the ramp sections, the non-slope roads of different heights connected front and back can be determined. Divide into segments.
  • the index values of the point cloud frame numbers belonging to the same slope area are continuous, all first point clouds can be classified according to the continuity between the index values of the point cloud frame numbers corresponding to the first point cloud data.
  • the point cloud frame number index value corresponding to the data is divided into multiple first index value ranges, where the number of first index value ranges is used to represent the number of times the target vehicle goes uphill or downhill. For example, as shown in Figure 1b, if there are 4 passes uphill or downhill, the index values saved in step S220 can be divided into 4 groups, such as [245,283], [996,1020], [1642,1666] and [1883,1910].
  • step S220 is to initially divide each frame of point cloud data into the first point cloud data corresponding to the ramp area and non- The second point cloud data corresponding to the ramp area. As shown in Figure 1b, outside the first index value range, there are still some point cloud data belonging to the slope area. Therefore, in order to make the point cloud frame data covered by each first index value range more comprehensive, the slope The point cloud data corresponding to the channel area is determined more accurately. For each first index value range corresponding to the first point cloud data, this embodiment can expand the processing.
  • a set number of index values can be added to both ends of each first index value range, thereby expanding the first index value range, obtaining an expanded target index value range, and adding each target index
  • the trajectories to be processed corresponding to the value range are respectively used as ramp areas.
  • index values that conform to the monotonic increase can be added at both ends of the first index value range to expand the first index value range. Get the expanded target index value range.
  • index values that conform to the monotonic decrease can be added at both ends of the first index value range to expand the first index value range, thereby obtaining range expansion.
  • the target index value range after.
  • index values that comply with the monotonicity are added at both ends of the index value range, so as to adjust the height information of the target vehicle to the first index value range.
  • the steps for expanding an index value range include:
  • Absolute value of height difference according to the monotonic increase between the absolute values of height difference when the vehicle is driving on the slope, determine the index values to be increased at both ends of the index value range, and according to the size relationship of the index values, add The increased index value is combined with the index value in the original index value range to form an expanded target index value range.
  • the index value corresponding to the middle part of the ramp can be summed by the endpoint values at both ends of the first index value range, and half of the sum value is used as the intermediate index value corresponding to the middle part of the ramp.
  • the intermediate index value After determining the intermediate index value, you can grow and extend from the middle part of the ramp to both sides, for example, extend forward from frame 264, that is, determine other index values that are smaller than the intermediate index value, such as frame 263, frame 262, Whether the height information of the target vehicle corresponding to the point cloud frames such as frame 261 and the height difference absolute value of frame 264 conforms to the monotonic increase property, and the index value that conforms to the independent increase property is regarded as the index value belonging to the target index value range. .
  • the absolute value of the height difference corresponding to the extended point cloud frame does not comply with the monotonic increase property compared to the height difference corresponding to the intermediate point cloud frame, it means that it extends to one end of the ramp.
  • the intermediate index value can be extended to the other end based on the intermediate index value.
  • other index values greater than the intermediate index value can be determined, such as frame 265, frame 266, frame 267, etc.
  • the height information of the target vehicle corresponding to these point cloud frames is relative to the height information of the target vehicle. Whether the absolute value of the height difference of the 264th frame conforms to the monotonically increasing property, and the index value that conforms to the independent increasing property is regarded as the index value belonging to the target index value range.
  • the height information corresponding to the extended point cloud frame does not comply with the monotonically increasing absolute value of the height difference corresponding to the intermediate point cloud frame, it is determined that it extends to the other end of the ramp.
  • the index values to be increased at both ends of the original first index value range can be determined, thereby combining the index values to be increased with the index values in the original first index value range to form a target index after the range is expanded.
  • the value range is to obtain the index value corresponding to a complete ramp area. For example, after expanding the range [245,283], the obtained range is [220,295].
  • the first point cloud data is divided according to the point cloud frame number index value.
  • the trajectory to be processed can be divided into multiple segments based on each target index value range, in order: non Slope section - slope section - non-slope section - slope section - non-slope section,..., and so on.
  • the number of slope segments determined in this embodiment represents the number of times the target vehicle goes uphill or downhill.
  • the point cloud frame number index value corresponding to the second point cloud data except the first point cloud data can be divided into a plurality of second index values. Range, that is, determine the index value range of point cloud data belonging to non-slope areas.
  • the target index value range obtained after expanding the first index value range is [220, 295], [971, 1034], based on the continuity between point cloud frame number index values, and based on the non-ramp area through
  • This principle of connecting ramp areas can divide the trajectories to be processed into: [0,219], [220,295], [296,970], [971,1034] (see Table 1 below for details).
  • the second index value range corresponding to the non-ramp sub-region is: [0,219] and [296,970].
  • step S260 For the specific hierarchical manner of step S260, please refer to the description of the above embodiments and will not be described again here.
  • the expanded target index value range can be covered
  • the point cloud frame data is more comprehensive, thereby improving the accuracy of ramp area division.
  • the expanded target index value range can correspond to a completed ramp, effectively improving the slope
  • the determination accuracy of the area is improved, thereby also improving the determination accuracy of the non-slope area.
  • FIG 3 is a flow chart of a point cloud data layering method provided in Embodiment 3 of the present invention. Based on the above embodiment, this embodiment provides a point cloud data application method, as shown in Figure 3 ,
  • steps S310 to S330 For the specific implementation of steps S310 to S330, reference may be made to the description of the above embodiment, and details will not be described again here.
  • the hierarchical results of the non-ramp area may include: the number of layers of the non-ramp area in the trajectory to be processed, the average height information corresponding to each layer of point cloud data, and the index value range of each layer of point cloud frame data.
  • the point cloud data of different layer heights can be visually displayed.
  • point cloud data at different heights can be represented by different colors; or point cloud data at different heights can also be represented by different shapes, which is not limited in this embodiment.
  • different workers can construct maps of point cloud data at different heights.
  • each layer of point cloud data can be used to create a map corresponding to the layer height
  • plane fitting can be performed on it based on a preset plane fitting algorithm to obtain the corresponding plane.
  • the preset plane fitting algorithm can be a plane fitting algorithm based on the plane extraction technology of RANSAC (Random Sample Consensus, random sampling consistency), or can also be a plane fitting algorithm based on the least squares method, etc.
  • RANSAC Random Sample Consensus, random sampling consistency
  • each frame of point cloud data can be projected into the pixel coordinate system, and the color of the projected pixel point is given to the point cloud to be colored, so as to color all the map elements in the plane, thereby obtaining each plane. corresponding map.
  • the ramp area can also be mapped according to the map creation method corresponding to the non-ramp area, thereby obtaining a map containing the ramp area and the non-ramp area in scenes such as parking lots.
  • the vehicle can be positioned based on the created high-precision map by using the layering of point cloud data provided by the embodiment of the present invention.
  • This method can layer the trajectories to be processed used in mapping and then layer the high-precision map to obtain the layer information of the current vehicle.
  • the technical solution provided by this embodiment uses the layered results of point cloud data to visually display point cloud data at different layer heights, and allows different workers to construct maps for different layers, thereby improving mapping efficiency.
  • the created map can be used in the vehicle positioning process to provide the self-driving vehicle with the layer information of the current road surface.
  • FIG. 4 is a structural block diagram of a point cloud data layering device provided in Embodiment 4 of the present invention. As shown in Figure 4, the device includes: a pitch angle determination module 410, a region division module 420 and a layering module 430, where ,
  • the pitch angle determination module 410 is configured to, for each frame of point cloud data corresponding to the trajectory to be processed, determine the vehicle pitch angle of the target vehicle relative to the horizontal plane when collecting each frame of point cloud data;
  • the area division module 420 is configured to divide the trajectory to be processed into a ramp area and a non-ramp area based on the size relationship between the vehicle pitch angles corresponding to each frame of point cloud data;
  • the layering module 430 is configured to layer the non-ramp areas according to height information.
  • the device provided by the embodiment of the present invention also includes:
  • the mapping module is configured to, after stratifying each non-ramp area according to height information, visually display the point cloud data in different layers according to the stratification results of the non-ramp area, and generate corresponding heights of different layers. planes to create maps corresponding to each plane.
  • area division module 420 includes:
  • the point cloud data dividing unit is configured to, in each frame of point cloud data, use the point cloud data corresponding to the pitch angle of the vehicle greater than the preset angle as the first point cloud of the target vehicle when it is going uphill or downhill. data;
  • the area dividing unit is configured to use the to-be-processed trajectories corresponding to the set of all first point cloud data as the ramp area, and to use the to-be-processed trajectories corresponding to the set of all second point cloud data except the first point cloud data. Treat tracks as non-ramp areas.
  • regional division units include:
  • the index value range determination subunit is configured to divide the point cloud frame number index values corresponding to all first point cloud data into multiple first index value ranges based on the continuity between point cloud frame number index values, where, The number of the first index value range is used to represent the number of times the target vehicle goes uphill or downhill;
  • the ramp area determination subunit is configured to expand the first index value range for any first index value range by adding several index values at both ends of the index value range to obtain the expanded range.
  • target index value range, and the trajectories to be processed corresponding to each target index value range are used as ramp areas;
  • the non-ramp area dividing subunit is configured to divide the point cloud corresponding to the second point cloud data except the first point cloud data based on the multiple target index value ranges corresponding to the first point cloud data and the continuity.
  • the frame number index value is divided into multiple second index value ranges, and the trajectories to be processed corresponding to each second index value range are respectively regarded as non-ramp areas.
  • ramp area determination subunits include:
  • the ramp area determination component is configured to, for any first index value range, according to the monotonicity between the target vehicle height information corresponding to each index value in the first index value range, at the two end positions of the index value range Index values that comply with the monotonicity are respectively added to expand the first index value range to obtain the expanded target index value range, and the to-be-processed trajectories corresponding to each target index value range are respectively used as ramp areas.
  • the monotonicity includes monotonic increasing property or monotonic decreasing property.
  • ramp area determination component specifically configured as:
  • the increased index value is combined with the index value in the original first index value range to form an expanded target index value range, and the to-be-processed trajectory corresponding to each target index value range is respectively as ramp areas.
  • the vehicle pitch angle is the vehicle pitch angle after filtering.
  • Optional layering module 430 includes:
  • the sub-region determination unit is configured to determine sub-regions with the same average height information from the non-ramp area, wherein the point cloud frame number index value corresponding to each sub-region is continuous;
  • the layering unit is configured to divide the non-ramp area into multiple layers in descending order of the average height information of different sub-areas.
  • the pitch angle determination module 410 is specifically configured:
  • the vehicle body coordinate system is a coordinate system that is fixedly connected to the vehicle body
  • the standard coordinate system is the coordinate system corresponding to the horizontal plane
  • the vehicle pitch angle of the target vehicle relative to the horizontal plane is determined when each frame of point cloud data is collected.
  • the multiple sensors include IMU, GPS, radar and/or image sensors.
  • the point cloud data layering device provided by the embodiment of the present invention can execute the point cloud data layering method provided by any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.
  • the hierarchical method of point cloud data provided by any embodiment of the present invention.
  • FIG. 5 is a structural block diagram of an electronic device provided in Embodiment 5 of the present invention. As shown in Figure 5, the electronic device includes:
  • Memory 510 storing executable program code
  • processor 520 coupled to memory 510;
  • the processor 520 calls the executable program code stored in the memory 510 to execute the point cloud data layering method provided by any embodiment of the present invention.
  • another embodiment of the present invention provides a vehicle, which includes the device as described in any of the above embodiments, or includes the electronic device as described above.
  • FIG. 6 is a schematic diagram of a vehicle provided in Embodiment 5 of the present invention.
  • the vehicle includes a speed sensor 61, an ECU (Electronic Control Unit) 62, a GPS (Global Positioning System) positioning device 63, and a T-Box (Telematics Box). 64.
  • the speed sensor 61 is used to measure the vehicle speed and uses the vehicle speed as the empirical speed for model training;
  • the GPS positioning device 63 is used to obtain the current geographical location of the vehicle;
  • the T-Box64 can be used as a gateway to communicate with the server;
  • the ECU62 can perform the above points A layered approach to cloud data.
  • the vehicle may also include: V2X (Vehicle-to-Everything, Internet of Vehicles) module 65 , radar 66 and camera 67 .
  • V2X Vehicle-to-Everything, Internet of Vehicles
  • the V2X module 65 is used to communicate with other vehicles, roadside equipment, etc.; the radar 66 or the camera 67 is used to sense the road environment information ahead and/or in other directions to obtain original point cloud data; the radar 66 and/or the camera 67 can be configured At the front and/or at the rear of the car.
  • another embodiment of the present invention provides a storage medium on which executable instructions are stored. When executed by a processor, the instructions enable the processor to implement point cloud data as described in any of the above embodiments. layered approach.
  • B corresponding to A means that B is associated with A, and B can be determined based on A.
  • determining B based on A does not mean determining B only based on A.
  • B can also be determined based on A and/or other information.
  • each functional unit in various embodiments of the present invention may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit.
  • the above integrated units can be implemented in the form of hardware or software functional units.
  • the integrated unit 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-accessible memory.
  • the technical solution of the present invention is essentially, or the part that contributes to the existing technology, or all or part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a memory , including several requests to cause a computer device (which can be a personal computer, a server or a network device, etc., specifically a processor in a computer device) to execute some or all of the steps of the above methods in various embodiments of the present invention.
  • a computer device which can be a personal computer, a server or a network device, etc., specifically a processor in a computer device
  • the program can be stored in a computer-readable storage medium, and the storage medium includes a read-only storage medium.
  • Memory Read-Only Memory, ROM), Random Access Memory (RAM), Programmable Read-only Memory (PROM), Erasable Programmable Read Only Memory, EPROM), One-time Programmable Read-Only Memory (OTPROM), Electronically Erasable Programmable Read-Only Memory (EEPROM), Compact Disc Read-Only Memory (CD-ROM) or other optical disk storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.
  • Memory Read-Only Memory, ROM
  • RAM Random Access Memory
  • PROM Programmable Read-only Memory
  • EPROM Erasable Programmable Read Only Memory
  • OTPROM One-time Programmable Read-Only Memory
  • EEPROM Electronically Erasable Programmable Read-Only Memory
  • CD-ROM Compact Disc Read-Only Memory
  • modules in the device in the embodiment may be distributed in the device in the embodiment according to the description of the embodiment, or may be correspondingly changed and located in one or more devices different from this embodiment.
  • the modules of the above embodiments can be combined into one module, or further divided into multiple sub-modules.

Landscapes

  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Computer Graphics (AREA)
  • Geometry (AREA)
  • Software Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Traffic Control Systems (AREA)
  • Length Measuring Devices By Optical Means (AREA)

Abstract

本发明实施例公开一种点云数据的分层方法、装置、设备、介质及车辆，其中，分层方法包括：对于待处理轨迹对应的各帧点云数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角；根据各帧点云数据分别对应的车辆俯仰角之间的大小关系，将待处理轨迹划分为坡道区域和非坡道区域；按照高度信息对非坡道区域进行分层。通过采用上述技术方案，避免了将不同层的点云数据划分为同一层点云数据的问题，提高了点云数据分层结果的准确性。

Description

一种点云数据的分层方法、装置、设备、介质及车辆 技术领域
本发明实施例涉及自动驾驶技术领域，具体而言，涉及一种点云数据的分层方法、装置、设备、介质及车辆。
背景技术
在制作用于辅助车辆定位技术的高精度地图时，尤其是在停车场特别是地下停车场这一场景下，需要将车辆在行驶过程中所采集的三维的点云数据进行分层处理，得到不同层高的点云数据，从而针对每一层高度的点云数据，制作对应的高精度地图。
相关技术中，主要是对车辆的行驶轨迹进行平面拟合和分割，并找出每一个平面当做一层。但由于在地下停车场或者是高架桥等场景下，同一层往往不是同一个平面，因此，上述方案很容易出现分层错误，从而影响后续高精度地图的制作。
发明内容
本发明实施例提供一种点云数据的分层方法、装置、设备、介质及车辆，用以克服点云数据分层不准确的问题。
具体技术方案如下：
第一方面，本发明实施例提供了一种点云数据的分层方法，包括：
对于待处理轨迹对应的各帧点云数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角；
根据各帧点云数据分别对应的车辆俯仰角之间的大小关系，将待处理轨迹划分为坡道区域和非坡道区域；
按照高度信息对非坡道区域进行分层。
通过上述方案可知，相对于相关技术中通过对待处理轨迹进行平面拟合和分割以对点云数据进行分层处理的方式，本发明实施例考虑到车辆在坡道路面和非坡道路面行驶时车辆姿态的差异性，通过确定用于表征车辆姿态的俯仰角，并根据车辆在非坡道路面行驶和在坡道路面行驶时车辆俯仰角的不同，将车辆行驶轨迹划分为坡道区域和非坡道区域，然后再对非坡道区域进行分层处理，避免了将不同层的点云数据划分为同一层点云数据的问题，提高了点云数据分层结果的准确性，从而有助于提高后续高精度地图的制作精度。
可选的，在按照高度信息对非坡道区域进行分层之后，本发明实施例提供的方法还包括：
按照非坡道区域的分层结果，对不同层中的点云数据进行可视化显示，并生成不同层高度对应的平面，以用于创建各平面对应的地图。
通过上述方案可知，对于点云数据的分层结果，可对不同层中的点云数据进行可视化显示，从而使得不同层分别由不同作业人员进行建图，以提高建图效率。对于创建完成的高精度地图，可将其应用于自动驾驶车辆的定位过程中，在高架桥、停车场，特别是地下停车场的场景下，可为自动驾驶车辆提供当前所在的路面的层数信息。
可选的，根据各帧点云数据分别对应的车辆俯仰角之间的大小关系，将所述待处理轨迹划分为坡道区域和非坡道区域，包括：
在各帧点云数据中，将车辆俯仰角大于预设角度的俯仰角所对应的点云数据，作为目标车辆在上坡或下坡时的第一点云数据；
将所有第一点云数据的集合所对应的待处理轨迹作为坡道区域，并将除第一点云数据之外的所有第二点云数据的集合所对应的待处理轨迹作为非坡道区域。
可选的，将所有第一点云数据的集合所对应的待处理轨迹作为坡道区域，并将除第一点云数据之外的所有第二点云数据的集合所对应的待处理轨迹作为非坡道区域，包括：
根据点云帧数索引值之间的连续性，将所有第一点云数据对应的点云帧数索引值划分为多个第一索引值范围，其中，第一索引值范围的个数用于表示目标车辆上坡或下坡的次数；
对于任意一个第一索引值范围，通过在该索引值范围两端位置分别增加若干个索引值，以将该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，并将各目标索引值范围所对应的待处理轨迹分别作为坡道区域；
基于第一点云数据对应的多个目标索引值范围以及点云帧数索引值之间的连续性，将除第一点云数据之外的第二点云数据对应的点云帧数索引值划分为多个第二索引值范围，并将各第二索引值范围所对应的待处理轨迹分别作为非坡道区域。
通过上述方案可知，基于车辆在坡道和非坡道区域行驶时车辆俯仰角的不同，本发明实施例根据每一帧点云数据对应的车辆俯仰角与预设角度之间的关系，将各帧点云数据初步划分为坡道区域对应的第一点云数据和非坡道区域对应的第二点云数据。其中，坡道区域可通过第一点云数据对应的第一索引值范围来表示。本发明实施例为了使得各个第一索引值范围所覆盖的点云帧数据更加全面，通过对第一点云数据对应的每个第一索引值范围进行扩大处理，可使得坡道区域对应的点云数据的确定更加准确。
可选的，通过在该索引值范围两端位置分别增加若干个索引值，以将该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，包括：
按照该第一索引值范围中各索引值所对应的目标车辆高度信息之间的单调性，在该索引值范围两端位置分别增加符合所述单调性的索引值，以对该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，其中，所述单调性包括单调递增性或单调递减性。
可选的，按照该第一索引值范围中各索引值所对应的目标车辆高度信息之间的单调性，在该索引值范围两端位置分别增加符合所述单调性的索引值，以对该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，包括：
从该第一索引值范围中，确定坡道中间部位对应的中间索引值；
确定所述中间索引值对应的目标车辆高度信息与该第一索引值范围中其他索引值对应的目标车辆高度信息之间的各高度差绝对值；
根据车辆在坡道行驶过程中各高度差绝对值之间的单调递增性，确定在该第一索引值 范围两端位置待增加的索引值；
按照索引值的大小关系，将待增加的索引值与原第一索引值范围中的索引值进行组合，以形成范围扩大后的目标索引值范围。
通过上述方案可知，本发明实施例根据目标车辆在坡道区域行驶时高度信息的单调递增性或单调递减性，在从初步划分的坡道区域对应的第一索引值范围中确定出与坡道中间部位对应的中间索引值后，通过确定出与中间索引值的高度信息之间的高程差符合该单调递增性的其他索引值，可对坡道区域对应的第一索引值的范围进行扩大处理，这样可使得扩大后的目标索引值范围能够对应一个完成的坡道区域，从而有效提高了坡道区域的确定精度。
可选的，车辆俯仰角为经过滤波处理后的车辆俯仰角。
通过上述方案可知，本发明实施例通过对车辆俯仰角进行滤波处理，解决车辆在行驶过程中存在微小抖动时所带来的干扰，提高了车辆俯仰角的准确性，从而使得后续坡道区域和非坡道区域的划分更加准确。
可选的，按照高度信息对非坡道区域划进行分层，包括：
从非坡道区域中确定平均高度信息相同的子区域，其中，每个子区域对应的点云帧数连续；
按照不同子区域的平均高度信息从大到小的顺序，将非坡道区域分为多层。
通过上述方案可知，考虑到高度不同的路面之间通常是通过坡道进行连通的，本发明实施例是先从待处理轨迹中确定出坡道区域，然后将待处理轨迹中剩下的区域确定为非坡道区域，并通过按照高度聚类算法将非坡道区域按照高度从高到低的顺序分为多层。相对于相关技术中对待处理轨迹进行平面拟合和分割的方式时，由于路面并非完全水平而容易导致将该路面拟合为多层的情况，本发明实施例在分层过程中，并不关注每层非坡道区域对应的路面是否水平，也无需考虑平面拟合时是否采用了其他平面的点云数据，本发明实施例提供的点云分层方案有效提高了点云数据分层的准确性。
可选的，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角，包括：
基于车体坐标系与标准坐标系之间的相对关系，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角，其中，车体坐标系为与车体固连的坐标系，标准坐标系为水平面对应的坐标系；或者，
基于惯性测量单元IMU采集的数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角；或者，
基于多传感器融合后的数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角，其中，多传感器包括IMU、GPS、雷达和/或图像传感器。
第二方面，本发明实施例还提供了一种点云数据的分层装置，包括：
俯仰角确定模块，被配置为对于待处理轨迹对应的各帧点云数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角，其中，所述标准坐标系用于表示水平面；
区域划分模块，被配置为根据各帧点云数据分别对应的车辆俯仰角之间的大小关系，将待处理轨迹划分为坡道区域和非坡道区域；
分层模块，被配置为按照高度信息对非坡道区域进行分层。
可选的，本发明实施例提供的装置还包括：
建图模块，被配置为在按照高度信息对非坡道区域进行分层之后，按照非坡道区域的分层结果，对不同层中的点云数据进行可视化显示，并生成不同层高度对应的平面，以用于创建各平面对应的地图。
可选的，区域划分模块，包括：
点云数据划分单元，被配置为在各帧点云数据中，将车辆俯仰角大于预设角度的俯仰角所对应的点云数据，作为目标车辆在上坡或下坡时的第一点云数据；
区域划分单元，被配置为将所有第一点云数据的集合所对应的待处理轨迹作为坡道区域，并将除第一点云数据之外的所有第二点云数据的集合所对应的待处理轨迹作为非坡道区域。
可选的，区域划分单元，包括：
索引值范围确定子单元，被配置为根据点云帧数索引值之间的连续性，将所有第一点云数据对应的点云帧数索引值划分为多个第一索引值范围，其中，第一索引值范围的个数用于表示目标车辆上坡或下坡的次数；
坡道区域确定子单元，被配置为对于任意一个第一索引值范围，通过在该索引值范围两端位置分别增加若干个索引值，以将该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，并将各目标索引值范围所对应的待处理轨迹分别作为坡道区域；
非坡道区域划分子单元，被配置为基于第一点云数据对应的多个目标索引值范围以及点云帧数索引值之间的连续性，将除第一点云数据之外的第二点云数据对应的点云帧数索引值划分为多个第二索引值范围，并将各第二索引值范围所对应的待处理轨迹分别作为非坡道区域。
可选的，坡道区域确定子单元，包括：
坡道区域确定组件，被配置为对于任意一个第一索引值范围，按照该第一索引值范围中各索引值所对应的目标车辆高度信息之间的单调性，在该索引值范围两端位置分别增加符合所述单调性的索引值，以对该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，并将各目标索引值范围所对应的待处理轨迹分别作为坡道区域；其中，单调性包括单调递增性或单调递减性。
可选的，坡道区域确定组件，具体被配置为：
对于任意一个第一索引值范围，从该第一索引值范围中，确定坡道中间部位对应的中间索引值；
确定中间索引值对应的目标车辆高度信息与该第一索引值范围中其他索引值对应的目标车辆高度信息之间的各高度差绝对值；
根据车辆在坡道行驶过程中各高度差绝对值之间的单调递增性，确定在该第一索引值范围两端位置待增加的索引值；
按照索引值的大小关系，将增加的索引值与原第一索引值范围中的索引值进行组合，以形成范围扩大后的目标索引值范围，并将各目标索引值范围所对应的待处理轨迹分别作为坡道区域。
可选的，车辆俯仰角为经过滤波处理后的车辆俯仰角。
可选的，分层模块，包括：
子区域确定单元，被配置为从非坡道区域中确定平均高度信息相同的子区域，其中， 每个子区域对应的点云帧数索引值连续；
分层单元，被配置为按照不同子区域的平均高度信息从大到小的顺序，将非坡道区域分为多层。
可选的，俯仰角确定模块，具体被配置：
基于车体坐标系与标准坐标系之间的相对关系，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角，其中，车体坐标系为与车体固连的坐标系，标准坐标系为水平面对应的坐标系；
基于惯性测量单元IMU采集的数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角；或者，
基于多传感器融合后的数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角，其中，多传感器包括IMU、GPS、雷达和/或图像传感器。
第三方面，本发明实施例提供了一种存储介质，其上存储有计算机程序，所述程序被处理器执行时实现如第一方面任一实施方式所述的方法。
第四方面，本发明实施例提供了一种电子设备，所述电子设备包括：
一个或多个处理器；
存储装置，用于存储一个或多个程序，
当所述一个或多个程序被所述一个或多个处理器执行，使得所述一个或多个处理器实现如第一方面任一实施方式所述的方法。
第五方面，本发明实施例提供了一种车辆，所述车辆包含如第二方面任一实施方式所述的装置，或者包含如第四方面所述的电子设备。
第六方面，本发明实施例提供了一种计算机程序，所述计算机程序包括程序指令，所述程序指令被计算机执行时实现如第一方面任一实施方式所述的方法。
附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1a为本发明实施例一提供的一种点云数据的分层方法的流程图；
图1b为本发明实施例一提供的车辆俯仰角与点云帧数索引值之间的关系示意图；
图1c为本发明实施例一提供的一种地库轨迹的侧视图；
图2为本发明实施例二提供的一种点云数据的分层方法的流程图；
图3为本发明实施例三提供的一种点云数据的分层方法的流程图；
图4为本发明实施例四提供的一种点云数据的分层装置的结构框图；
图5为本发明实施例五提供的一种电子设备的结构框图；
图6为本发明实施例五提供的一种车辆的示意图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有付出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
需要说明的是，本发明实施例及附图中的术语“包括”和“具有”以及它们任何变形，意图在于覆盖不排他的包含。例如包含了一系列步骤或单元的过程、方法、系统、产品或设备没有限定于已列出的步骤或单元，而是可选地还包括没有列出的步骤或单元，或可选地还包括对于这些过程、方法、产品或设备固有的其它步骤或单元。
本发明实施例公开了一种点云数据的分层方法、装置、设备、介质及车辆。以下分别进行详细说明。
实施例一
图1a为本发明实施例一提供的一种点云数据的分层方法的流程图，该方法可应用于车载电脑、车载工业控制计算机(Industrial personal Computer，IPC)等车载终端，也可应用于服务器，本发明实施例对此不做限定。本实施例提供的方法可应用于高架桥、停车场，尤其是地下停车场等场景下点云数据的分层处理过程中，其分层结果可用于地图创建或者车辆定位等。本实施例提供的方法可由点云数据的分层装置来执行，该装置可通过软件和/或硬件的方式实现。如图1a所示，本实施例提供的方法具体包括：
S110、对于待处理轨迹对应的各帧点云数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角。
其中，待处理轨迹为目标车辆在目标场景中行驶生成的轨迹，该待处理轨迹包括按照时间先后顺序排序的多个轨迹点。点云数据为目标车辆在每一轨迹点处观测到的目标场景对应的点云数据。其中，目标场景可以为高架桥、停车场，特别是地下停车场等场景。目标车辆可以为安装有多种传感器设备的地图采集车。
示例性的，点云数据的获取方式有多种，例如，可通过在目标车辆上设置的测距传感器(例如雷达或激光扫描仪等)采集得到；或者，也可利用在目标车辆上所设置的图像采集装置(例如深度摄像头或双目摄像头等)采集每一轨迹点的图像，再基于各轨迹点的图像得到各轨迹点的点云数据。本发明实施例对点云数据的具体获取方式不进行限定，凡是能够获得每一轨迹点处观测到的目标场景对应的点云数据的方式均可以应用于本发明实施例中。
可以理解的是，车辆在坡道路段和非坡道路段行驶时车辆姿态存在明显的不同。当车辆在非坡道路段行驶时，车辆姿态和水平面是平行的，当车道在坡道路段行驶时，车辆姿态和水平面之间会存在一定的夹角。本实施例中，该夹角可通过车辆姿态的俯仰角来表示。
本实施例中，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角的方式有多种，作为一种可选的实施方式，可先选定一个表征水平面的标准坐标系，该标准坐标系可以是以当地水平面上的一点为原点，以水平面为xoy平面所建立的坐标系。具体的，该标准坐标系可以为ENU坐标系(东北天坐标系)。本实施例中，可根据车体坐标系与标准坐标系之间的相对关系，得到车辆相对于标准坐标系的初始俯仰角。其中，车体坐标系定义在车体上，是与车体固连的坐标系，例如可定义在车体后轴的中心位置，如果车辆位于水平面，则车体坐标系也处于水平状态。
作为另一种可选的实施方式，由于惯性测量单元(Inertial Measurement Unit，IMU)可感知重力方向，因此可通过IMU确定目标车辆在采集各帧点云数据时相对于水平面的车辆姿态的初始俯仰角。
作为另一种可选的实施方式，可基于多传感器融合后的数据，确定车辆的姿态，从而确定该姿态相对于水平面的初始车辆俯仰角。其中，多传感器可以包括IMU、GPS(Global Positioning System，全球定位系统)、雷达和/或图像传感器。
进一步的，在得到车辆姿态的初始俯仰角之后，可对初始俯仰角进行滤波处理，以解决车辆在行驶过程中存在微小抖动时所带来的干扰，从而提高车辆俯仰角的准确性。
示例性的，滤波处理的方法有多种，例如可以采用均值滤波方式，具体为：对于当前帧点云数据对应的初始车辆俯仰角，通过该帧点云数据前后两帧的点云数据所对应的车辆俯仰角对当前帧点云数据对应的车辆俯仰角进行均值滤波处理，并将均值滤波处理后得到的俯仰角数值作为当前帧点云数据对应的车辆俯仰角，该均值滤波处理过程可通过如下公式来表示：
p[i] (后)＝(p[i-2]+p[i-1]+p[i] (前)+p[i+1]+p[i+2])/5；
其中，i表示当前帧点云数据的索引值，p[i] (前)表示滤波处理前当前帧点云数据对应的车辆俯仰角，p[i] (后)表示滤波处理后前当前帧点云数据对应的车辆俯仰角。
除上述均值滤波之外，还可以采用高斯滤波或中值滤波等方式对每一帧点云数据对应的车辆俯仰角进行滤波处理，本实施例对滤波方式不作具体限定。
本实施例中，基于目标车辆在非坡道路段和坡道路段行驶时车辆俯仰角的不同，可确定出目标车辆行驶的是坡道路段还是非坡道路段。图1b为本发明实施例一提供的车辆俯仰角与点云帧数索引值之间的关系示意图，如图1b所示，横坐标表示的是点云帧数索引值，纵坐标表示的是车辆俯仰角。图中所框出的四个尖峰区域A的俯仰角的数值变化范围较大，对应的是车辆在坡道路段行驶时的俯仰角。其中，坡道路段可以是车辆从高度较高一层路段行驶到较低的一层路段，也可以是从高度较低的一层路段行驶到较高的一层路段。图1b中所框出的区域B的俯仰角的数值变化范围较小，对应的是非坡道路段。由于坡道路段对应的车辆俯仰角在数值较大的范围内变化，非坡道路段对应的车辆俯仰角在数值较小的范围内变化，因此，本实施例中通过计算目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角，可确定出坡道路段对应的行驶轨迹，即找出如图1b所示的尖峰区域A对应的轨迹，进而可确定出待处理轨迹中非坡道区域对应的轨迹，即如图1b所示的区域B对应的轨迹。
S120、根据各帧点云数据分别对应的车辆俯仰角之间的大小关系，将待处理轨迹划分为坡道区域和非坡道区域。
可以理解的是，当车辆在高度不同的路段行驶时，由于不同路段之间存在一定的高程差，因此不同路段之间通常会经过斜坡进行过度，即高度不同的路段之间通常是通过坡道进行连通的。例如，图1c为本发明实施例一提供的一种地库轨迹的侧视图。如图1c所示，当车辆从地上路面驶入地下B1层车库时，或者从地下B1层车库驶入地上路面，或者通过地下B1层车库驶入地下B2层车库时，都会经过一段斜坡路段。当车辆在非坡道路段行驶时，车辆姿态和水平面是平行的，车辆俯仰角通常在取值较小的数值范围内变化。当车道开始驶入坡道路段，到车辆在坡道路段行驶，再到车辆驶出坡道路段的过程中，车辆 俯仰角通常会在取值较大的数值范围内变化。基于上述原理，可根据各帧点云数据分别对应的车辆俯仰角之间的大小关系，确定车辆行驶的是坡道区域还是非坡道区域。
作为一种可选的实施方式，可以预先设定一个角度，作为车辆在非坡道区域和坡道区域的临界角度。其中，该预设角度可根据车辆多次在非坡道路段和坡道路段行驶时俯仰角的经验值来确定。例如，当车辆在跨层行驶时，如图1b所示，车辆俯仰角通常会在5°到10°之间进行变化，本实施例中，可将预设角度设置为5°。
本实施例中，可将车辆俯仰角大于预设角度的俯仰角所对应的点云数据，作为目标车辆在上坡或下坡时的第一点云数据。由于不同高度的非坡道区域是通过坡道区域来连接的，并且由于待处理轨迹的连续性，在确定出坡道区域对应的第一点云数据后，可将除第一点云数据之外的所有第二点云数据的集合所对应的待处理轨迹作为非坡道区域。
S130、按照高度信息对非坡道区域进行分层。
本实施例中，由于待处理轨迹包括按照时间先后顺序排序的多个轨迹点，因此，在根据车辆俯仰角确定出坡道区域后，可根据时间顺序，将待处理轨迹划分为多段子区域，即划分为非坡道子区域、坡道子区域、非坡道子区域和坡道子区域等，其中，每个子区域对应的点云帧数连续。当车辆在高度不同的路段来回行驶时，例如车辆从高度较高一层的路段行驶到较低一层路段，再从较低一层路段再次行驶到较高一层路段时，待处理轨迹中的非坡道区域将包含多段高度值相同的非坡道子区域。
作为一种可选的实施方式，可根据非坡道子区域的平均高度值，将平均高度值在设定范围内的非坡道子区域作为同一层非坡道路面区域，从而可将非坡道区域划分为多层。其中，非坡道子区域的平均高度值的可根据每一点云帧对应的车辆高度信息来确定。其中，车辆高度信息可通过GPS来获取。
具体的，可按照每个子区域点云帧数的连续性，将非坡道区域划分为多个非坡道子区域。通过计算各非坡道子区域的平均高度，并对各非坡道子区域按照平均高度进行聚类，可得到高度信息相同的非坡道子区域，即可将车辆在不同时间段行驶过的同一高度的路段归为一类。其中，聚类高度可以设置为2米。在完成同一高度的非坡道子区域的聚类后，可按照不同非坡道子区域的平均高度信息从大到小的顺序，可将非坡道区域分为多层。相对于相关技术中对待处理轨迹进行平面拟合和分割以对点云数据进行分层时，由于路面并非完全水平而容易导致将该路面拟合为多层的情况，本实施例的分层过程，并不关注每层非坡道区域对应的路面是否水平，也无需考虑平面拟合时是否采用了其他平面的点云数据，本实施例提供的点云分层方案有效提高了点云数据分层的准确性。
本实施例中，考虑到车辆在坡道路面和非坡道路面行驶时车辆姿态的差异性，通过确定用于表征车辆姿态的俯仰角，并根据车辆在非坡道路段行驶和在坡道路段行驶时车辆俯仰角的不同，可将待处理轨迹划分为坡道区域和非坡道区域，然后再对非坡道区域进行分层处理。相对于相关技术中通过对待处理轨迹进行平面拟合和分割以对点云数据进行分层处理的方式，本实施例提供的技术方案避免了将不同层的点云数据划分为同一层点云数据的问题，提高了点云数据分层的准确性，从而有助于提高后续高精度地图的制作精度。
实施例二
图2为本发明实施例二提供的一种点云数据的分层方法的流程图，本实施例在上述实施例的基础上，对坡道区域和非坡道区域的具体划分过程进行了细化，如图2所示，本 实施例提供的方法包括：
S210、对于待处理轨迹对应的各帧点云数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角。
其中，步骤S210的具体实施方式可参照上述实施例的说明，此处不再赘述。
S220、在各帧点云数据中，将车辆俯仰角大于预设角度的俯仰角所对应的点云数据，作为目标车辆在上坡或下坡时的第一点云数据。
本实施例中，在确定出目标车辆在上坡或下坡时的第一点云数据，和除第一点云数据之外的非坡道区域对应的第二点云数据时，可将第一点云数据对应的点云帧数索引值，以及第二点云数据对应的点云帧数索引值进行存储。
S230、根据第一点云数据所对应的点云帧数索引值之间的连续性，将所有第一点云数据对应的点云帧数索引值划分为多个第一索引值范围。
本实施例中，考虑到不同高度的非坡道路段是通过坡道路段进行连通的，因此，可先对坡道路段进行确定，然后再基于坡道路段对前后连通的不同高度的非坡道路段进行划分。
具体的，由于属于同一坡道区域的点云帧数索引值是连续的，因此，可根据第一点云数据所对应的点云帧数索引值之间的连续性，将所有第一点云数据对应的点云帧数索引值划分为多个第一索引值范围，其中，第一索引值范围的个数用于表示目标车辆上坡或下坡的次数。例如，如图1b所示，如果有4次经过上坡或下坡，则可将步骤S220中保存的索引值分成4组，例如[245,283]，[996,1020]，[1642,1666]和[1883,1910]。
S240、对于任意一个第一索引值范围，通过在该索引值范围两端位置分别增加若干个索引值，以将该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，并将各目标索引值范围所对应的待处理轨迹分别作为坡道区域。
本实施例中，步骤S220是根据每一帧点云数据对应的车辆俯仰角与预设角度之间的关系，将各帧点云数据初步划分为坡道区域对应的第一点云数据和非坡道区域对应的第二点云数据。如图1b所示，在第一索引值范围外，仍存在一些属于该坡道区域的点云数据，因此，为了使得各个第一索引值范围所覆盖的点云帧数据更加全面，以使得坡道区域对应的点云数据的确定更加准确，对于第一点云数据对应的每个第一索引值范围，本实施例可将其进行扩大处理。
示例性的，可将各个第一索引值范围两端位置分别增加设定数目的索引值，从而将该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，并将各目标索引值范围所对应的待处理轨迹分别作为坡道区域。
示例性的，可基于目标车辆在上坡时车辆高度的单调递增性，在第一索引值范围两端位置分别增加符合该单调递增性的索引值，以对该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围。或者，可基于目标车辆在下坡时车辆高度的单调递减性，在第一索引值范围两端位置分别增加符合该单调递减性的索引值，以对该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围。
本实施例中，按照该第一索引值范围中各索引值所对应的目标车辆高度信息之间的单调性，在该索引值范围两端位置分别增加符合单调性的索引值，以对该第一索引值范围进行扩大的步骤包括：
从该第一索引值范围中，确定坡道中间部位对应的中间索引值，并确定中间索引值对 应的目标车辆高度信息与该索引值范围中其他索引值对应的目标车辆高度信息之间的各高度差绝对值；根据车辆在坡道行驶过程中各高度差绝对值之间的单调递增性，确定在该索引值范围两端位置待增加的索引值，并按照索引值的大小关系，将待增加的索引值与原索引值范围中的索引值进行组合，以形成范围扩大后的目标索引值范围。
具体的，坡道中间部位对应的索引值可通过第一索引值范围两端的端点值作和，并将和值的一半作为坡道中间部位对应的中间索引值。例如，对于上述第一索引值范围[245,283]，其对应的中间索引值为：(245+283)/2＝264。在确定出中间索引值之后，可分别从坡道的中间部位向两侧生长延伸，例如从第264帧向前延伸，即判断小于中间索引值的其他索引值，例如263帧、第262帧、第261帧等这些点云帧对应的目标车辆的高度信息相对于第264帧的高度差绝对值是否符合单调递增性，并将符合单独递增性的索引值作为属于目标索引值范围中的索引值。当延伸到的点云帧对应的高度信息相对于中间点云帧对应的高度差绝对值不符合单调递增性时，则说明延伸到了坡道的一端。此时，可基于中间索引值再向另一端延伸，例如判断大于中间索引值的其他索引值，例如265帧、第266帧、第267帧等这些点云帧对应的目标车辆的高度信息相对于第264帧的高度差绝对值是否符合单调递增性，并将符合单独递增性的索引值作为属于目标索引值范围中的索引值。当延伸到的点云帧对应的高度信息相对于中间点云帧对应的高度差绝对值不符合单调递增性时，则确定延伸到了坡道的另一端。按照上述方式，可确定出在原第一索引值范围两端位置待增加的索引值，从而将待增加的索引值与原第一索引值范围中的索引值进行组合，形成范围扩大后的目标索引值范围，即得到一个完整的坡道区域对应的索引值，例如将范围[245,283]扩大后，得到的范围是[220,295]。
S250、基于第一点云数据对应的多个目标索引值范围，将除第一点云数据之外的第二点云数据对应的点云帧数索引值划分为多个第二索引值范围，并将各第二索引值范围所对应的待处理轨迹分别作为非坡道区域。
由于目标车辆在行驶过程中可能会经过多次上坡或下坡路段，因此对于坡道区域对应的所有第一点云数据，在将该第一点云数据按照点云帧数索引值之间的连续性划分为多组后，即确定出待处理轨迹中的多个坡道路段分别对应的目标索引值范围后，可基于每个目标索引值范围将待处理轨迹划分为多段，依次为：非坡道路段—坡道路段—非坡道路段—坡道路段—非坡道路段，……，以此类推。其中，本实施例所确定的坡道路段的个数表示的是目标车辆上坡或下坡的次数。
本实施例中，可基于点云帧数索引值之间的连续性，将除第一点云数据之外的第二点云数据对应的点云帧数索引值划分为多个第二索引值范围，即对属于非坡道区域的点云数据的索引值范围进行确定。
具体的，对于将第一索引值范围扩大后得到的目标索引值范围为[220,295]、[971,1034]，基于点云帧数索引值之间的连续性，并且基于非坡道区域是通过坡道区域进行连通的这一原理，可将待处理轨迹划分为：[0,219]、[220,295]、[296,970]、[971,1034](具体请参见下表1)。其中，非坡道子区域对应的第二索引值范围为：[0,219]和[296,970]。
表1区域划分表
非坡道子区域 坡道子区域 非坡道子区域 坡道子区域 ……
[0,219] [220,295] [296,970] [971,1034] ……
S260、按照高度信息对非坡道区域进行分层。
其中，步骤S260的具体分层方式可参见上述实施例的说明，此处不再赘述。
本实施例中，对于按照车辆俯仰角初步划分得到的坡道区域对应的第一索引值范围，通过对该第一索引值范围进行扩大处理，可使得扩大后得到的目标索引值范围所覆盖的点云帧数据更加全面，从而提高了坡道区域划分的准确性。在进行扩大处理时，通过确定符合车辆在坡道区域行驶时车辆高度信息单调性的所有点云帧数据，可使得扩大后的目标索引值范围能够对应一个完成的坡道，有效提高了坡道区域的确定精度，从而也提高了非坡道区域的确定精度。
实施例三
图3为本发明实施例三提供的一种点云数据的分层方法的流程图，本实施例在上述实施例的基础上，提供了一种点云数据的应用方法，如图3所示，
S310、对于待处理轨迹对应的各帧点云数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角。
S320、根据各帧点云数据分别对应的车辆俯仰角之间的大小关系，将待处理轨迹划分为坡道区域和非坡道区域。
S330、按照高度信息对非坡道区域进行分层。
其中，步骤S310～S330的具体实现方式可参照上述实施例的说明，此处不再赘述。
S340、按照非坡道区域的分层结果，对不同层中的点云数据进行可视化显示，并生成不同层高度对应的平面，以用于创建各平面对应的地图。
其中，非坡道区域的分层结果可包括：待处理轨迹中非坡道区域的层数，每一层点云数据对应的平均高度信息，以及每一层点云帧数据的索引值范围。
本实施例中，为了对不同层高的点云数据进行有效区分，可对不同层高的点云数据进行可视化显示。具体可以通过不同的颜色来表示处于不同层高的点云数据；或者，也可通过不同的形状来表示不同层高的点云数据，本实施例对此不作限定。在创建地图时，为了提高生产效率，可通过不同的作业人员对不同层高的点云数据分别进行建图。
具体的，在利用各层点云数据创建对应层高的地图时，对于每一层点云数据，可以基于预设平面拟合算法对其进行平面拟合，得到对应的平面。其中，预设平面拟合算法可以为基于RANSAC(Random Sample Consensus，随机采样一致性)的平面提取技术的平面拟合算法，还可以为基于最小二乘法的平面拟合算法等。在完成平面拟合后，可为平面中的点进行着色。具体的，可将各帧点云数据投影到像素坐标系下，并将投影得到的像素点的颜色付给待着色的点云，以实现对平面中所有的地图元素进行着色，从而得到各平面对应的地图。类似的，坡道区域也可按照非坡道区域对应的地图创建方式进行建图，从而得到在停车场等场景下包含有坡道区域和非坡道区域的地图。
进一步的，在自动驾驶车辆的行驶过程中，在高架桥、停车场等应用场景下，基于已创建完成的高精度地图可对车辆进行定位，通过采用本发明实施例提供的点云数据的分层方法，可将建图使用的待处理轨迹进行分层进而可将高精地图分层，得到当前车辆所处的层数信息。
本实施例提供的技术方案，通过利用点云数据的分层结果，对不同层高的点云数据进行可视化显示，并将不同层由不同作业人员进行建图，提高了建图效率。其中，创建完成 的地图可用于车辆的定位过程中，为自动驾驶车辆提供当前所在的路面的层数信息。
实施例四
图4为本发明实施例四提供的一种点云数据的分层装置的结构框图，如图4所示，该装置包括：俯仰角确定模块410、区域划分模块420和分层模块430，其中，
俯仰角确定模块410，被配置为对于待处理轨迹对应的各帧点云数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角；
区域划分模块420，被配置为根据各帧点云数据分别对应的车辆俯仰角之间的大小关系，将待处理轨迹划分为坡道区域和非坡道区域；
分层模块430，被配置为按照高度信息对非坡道区域进行分层。
可选的，本发明实施例提供的装置还包括：
建图模块，被配置为在按照高度信息对各非坡道区域进行分层之后，按照非坡道区域的分层结果，对不同层中的点云数据进行可视化显示，并生成不同层高度对应的平面，以用于创建各平面对应的地图。
可选的，区域划分模块420，包括：
点云数据划分单元，被配置为在各帧点云数据中，将车辆俯仰角大于预设角度的俯仰角所对应的点云数据，作为目标车辆在上坡或下坡时的第一点云数据；
区域划分单元，被配置为将所有第一点云数据的集合所对应的待处理轨迹作为坡道区域，并将除第一点云数据之外的所有第二点云数据的集合所对应的待处理轨迹作为非坡道区域。
可选的，区域划分单元，包括：
索引值范围确定子单元，被配置为根据点云帧数索引值之间的连续性，将所有第一点云数据对应的点云帧数索引值划分为多个第一索引值范围，其中，第一索引值范围的个数用于表示目标车辆上坡或下坡的次数；
坡道区域确定子单元，被配置为对于任意一个第一索引值范围，通过在该索引值范围两端位置分别增加若干个索引值，以将该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，并将各目标索引值范围所对应的待处理轨迹分别作为坡道区域；
非坡道区域划分子单元，被配置为基于第一点云数据对应的多个目标索引值范围以及所述连续性，将除第一点云数据之外的第二点云数据对应的点云帧数索引值划分为多个第二索引值范围，并将各第二索引值范围所对应的待处理轨迹分别作为非坡道区域。
可选的，坡道区域确定子单元，包括：
坡道区域确定组件，被配置为对于任意一个第一索引值范围，按照该第一索引值范围中各索引值所对应的目标车辆高度信息之间的单调性，在该索引值范围两端位置分别增加符合所述单调性的索引值，以对该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，并将各目标索引值范围所对应的待处理轨迹分别作为坡道区域；其中，所述单调性包括单调递增性或单调递减性。
可选的，坡道区域确定组件，具体被配置为：
对于任意一个第一索引值范围，从该第一索引值范围中，确定坡道中间部位对应的中间索引值；
确定所述中间索引值对应的目标车辆高度信息与该第一索引值范围中其他索引值对 应的目标车辆高度信息之间的各高度差绝对值；
根据车辆在坡道行驶过程中各高度差绝对值之间的单调递增性，确定在该第一索引值范围两端位置待增加的索引值；
按照索引值的大小关系，将增加的索引值与原第一索引值范围中的索引值进行组合，以形成范围扩大后的目标索引值范围，并将各目标索引值范围所对应的待处理轨迹分别作为坡道区域。
可选的，车辆俯仰角为经过滤波处理后的车辆俯仰角。
可选的，分层模块430，包括：
子区域确定单元，被配置为从非坡道区域中确定平均高度信息相同的子区域，其中，每个子区域对应的点云帧数索引值连续；
分层单元，被配置为按照不同子区域的平均高度信息从大到小的顺序，将非坡道区域分为多层。
可选的，俯仰角确定模块410，具体被配置：
基于车体坐标系与标准坐标系之间的相对关系，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角，其中，车体坐标系为与车体固连的坐标系，标准坐标系为水平面对应的坐标系；或者，
基于惯性测量单元IMU采集的数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角；或者，
基于多传感器融合后的数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角，其中，多传感器包括IMU、GPS、雷达和/或图像传感器。
本发明实施例所提供的点云数据的分层装置可执行本发明任意实施例所提供的点云数据的分层方法，具备执行方法相应的功能模块和有益效果。未在上述实施例中详尽描述的技术细节，可参见本发明任意实施例所提供的点云数据的分层方法。
实施例五
图5为本发明实施例五提供的一种电子设备的结构框图，如图5所示，该电子设备包括：
存储有可执行程序代码的存储器510；
与存储器510耦合的处理器520；
其中，处理器520调用存储器510中存储的可执行程序代码，执行本发明任意实施例所提供的点云数据的分层方法。
基于上述实施例，本发明的另一实施例提供了一种车辆，所述车辆包含如上述任一实施例所述的装置，或者包含如上所述的电子设备。
图6为本发明实施例五提供的一种车辆的示意图。如图6所示，车辆包括速度传感器61、ECU(Electronic Control Unit，电子控制单元)62、GPS(Global Positioning System，全球定位系统)定位设备63、T-Box(Telematics Box，远程信息处理器)64。其中，速度传感器61用于测量车速，并将车速作为经验速度供模型训练使用；GPS定位设备63用于获取车辆的当前地理位置；T-Box64可以作为网关与服务器进行通信；ECU62可以执行上述点云数据的分层方法。
此外，该车辆还可以包括：V2X(Vehicle-to-Everything，车联网)模块65、雷达66 和摄像头67。V2X模块65用于与其他车辆、路侧设备等进行通信；雷达66或摄像头67用于感知前方和/或其他方向的道路环境信息，得到原始点云数据；雷达66和/或摄像头67可以配置在车身前部和/或车身尾部。
基于上述方法实施例，本发明的另一实施例提供了一种存储介质，其上存储有可执行指令，该指令被处理器执行时使处理器实现如上任一实施方式所述的点云数据的分层方法。
在本发明的各种实施例中，应理解，上述各过程的序号的大小并不意味着执行顺序的必然先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本发明实施例的实施过程构成任何限定。
在本发明所提供的实施例中，应理解，“与A相应的B”表示B与A相关联，根据A可以确定B。但还应理解，根据A确定B并不意味着仅仅根据A确定B，还可以根据A和/或其他信息确定B。
另外，在本发明各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。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
上述集成的单元若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获取的存储器中。基于这样的理解，本发明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全部或者部分，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器中，包括若干请求用以使得一台计算机设备(可以为个人计算机、服务器或者网络设备等，具体可以是计算机设备中的处理器)执行本发明的各个实施例上述方法的部分或全部步骤。
本领域普通技术人员可以理解上述实施例的各种方法中的全部或部分步骤是可以通过程序来指令相关的硬件来完成，该程序可以存储于一计算机可读存储介质中，存储介质包括只读存储器(Read-Only Memory，ROM)、随机存储器(Random Access Memory，RAM)、可编程只读存储器(Programmable Read-only Memory，PROM)、可擦除可编程只读存储器(Erasable Programmable Read Only Memory，EPROM)、一次可编程只读存储器(One-time Programmable Read-Only Memory，OTPROM)、电子抹除式可复写只读存储器(Electrically-Erasable Programmable Read-Only Memory，EEPROM)、只读光盘(Compact Disc Read-Only Memory，CD-ROM)或其他光盘存储器、磁盘存储器、磁带存储器、或者能够用于携带或存储数据的计算机可读的任何其他介质。
本领域普通技术人员可以理解：附图只是一个实施例的示意图，附图中的模块或流程并不一定是实施本发明所必须的。
本领域普通技术人员可以理解：实施例中的装置中的模块可以按照实施例描述分布于实施例的装置中，也可以进行相应变化位于不同于本实施例的一个或多个装置中。上述实施例的模块可以合并为一个模块，也可以进一步拆分成多个子模块。
最后应说明的是：以上实施例仅用以说明本发明的技术方案，而非对其限制；尽管参照前述实施例对本发明进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本发明实施例技术方案的精神和范围。

Claims (20)

  1. 一种点云数据的分层方法，其特征在于，包括：
    对于待处理轨迹对应的各帧点云数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角；
    根据各帧点云数据分别对应的车辆俯仰角之间的大小关系，将所述待处理轨迹划分为坡道区域和非坡道区域；
    按照高度信息对非坡道区域进行分层。
  2. 根据权利要求1所述的方法，其特征在于，在按照高度信息对非坡道区域进行分层之后，所述方法还包括：
    按照非坡道区域的分层结果，对不同层中的点云数据进行可视化显示，并生成不同层高度对应的平面，以用于创建各平面对应的地图。
  3. 根据权利要求1所述的方法，其特征在于，根据各帧点云数据分别对应的车辆俯仰角之间的大小关系，将所述待处理轨迹划分为坡道区域和非坡道区域，包括：
    在各帧点云数据中，将车辆俯仰角大于预设角度的俯仰角所对应的点云数据，作为目标车辆在上坡或下坡时的第一点云数据；
    将所有第一点云数据的集合所对应的待处理轨迹作为坡道区域，并将除第一点云数据之外的所有第二点云数据的集合所对应的待处理轨迹作为非坡道区域。
  4. 根据权利要求3所述的方法，其特征在于，将所有第一点云数据的集合所对应的待处理轨迹作为坡道区域，并将除第一点云数据之外的所有第二点云数据的集合所对应的待处理轨迹作为非坡道区域，包括：
    根据点云帧数索引值之间的连续性，将所有第一点云数据对应的点云帧数索引值划分为多个第一索引值范围，其中，第一索引值范围的个数用于表示目标车辆上坡或下坡的次数；
    对于任意一个第一索引值范围，通过在该索引值范围两端位置分别增加若干个索引值，以将该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，并将各目标索引值范围所对应的待处理轨迹分别作为坡道区域；
    基于第一点云数据对应的多个目标索引值范围以及所述连续性，将除第一点云数据之外的第二点云数据对应的点云帧数索引值划分为多个第二索引值范围，并将各第二索引值范围所对应的待处理轨迹分别作为非坡道区域。
  5. 根据权利要求4所述的方法，其特征在于，通过在该索引值范围两端位置分别增加若干个索引值，以将该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，包括：
    按照该第一索引值范围中各索引值所对应的目标车辆高度信息之间的单调性，在该索引值范围两端位置分别增加符合所述单调性的索引值，以对该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，其中，所述单调性包括单调递增性或单调递减性。
  6. 根据权利要求5所述的方法，其特征在于，按照该第一索引值范围中各索引值所对应的目标车辆高度信息之间的单调性，在该索引值范围两端位置分别增加符合所述单调性的索引值，以对该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，包括：
    从该第一索引值范围中，确定坡道中间部位对应的中间索引值；
    确定所述中间索引值对应的目标车辆高度信息与该第一索引值范围中其他索引值对应的目标车辆高度信息之间的各高度差绝对值；
    根据车辆在坡道行驶过程中各高度差绝对值之间的单调递增性，确定在该第一索引值范围两端位置待增加的索引值；
    按照索引值的大小关系，将待增加的索引值与原第一索引值范围中的索引值进行组合，以形成范围扩大后的目标索引值范围。
  7. 根据权利要求3所述的方法，其特征在于，所述车辆俯仰角为经过滤波处理后的车辆俯仰角。
  8. 根据权利要求1-7任一所述的方法，其特征在于，所述按照高度信息对非坡道区域划进行分层，包括：
    从所述非坡道区域中确定平均高度信息相同的子区域，其中，每个子区域对应的点云帧数索引值连续；
    按照不同子区域的平均高度信息从大到小的顺序，将所述非坡道区域分为多层。
  9. 根据权利要求1所述的方法，其特征在于，所述确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角，包括：
    基于车体坐标系与标准坐标系之间的相对关系，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角，其中，所述车体坐标系为与车体固连的坐标系，所述标准坐标系为水平面对应的坐标系；或者，
    基于惯性测量单元IMU采集的数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角；或者，
    基于多传感器融合后的数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角，其中，所述多传感器包括IMU、GPS、雷达和/或图像传感器。
  10. 一种点云数据的分层装置，其特征在于，包括：
    俯仰角确定模块，被配置为对于待处理轨迹对应的各帧点云数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角；
    区域划分模块，被配置为根据各帧点云数据分别对应的车辆俯仰角之间的大小关系，将所述待处理轨迹划分为坡道区域和非坡道区域；
    分层模块，被配置为按照高度信息对非坡道区域进行分层。
  11. 根据权利要求10所述的装置，其特征在于，所述装置还包括：
    建图模块，被配置为在按照高度信息对非坡道区域进行分层之后，按照非坡道区域的分层结果，对不同层中的点云数据进行可视化显示，并生成不同层高度对应的平面，以用于创建各平面对应的地图。
  12. 根据权利要求10所述的装置，其特征在于，所述区域划分模块，包括：
    点云数据划分单元，被配置为在各帧点云数据中，将车辆俯仰角大于预设角度的俯仰角所对应的点云数据，作为目标车辆在上坡或下坡时的第一点云数据；
    区域划分单元，被配置为将所有第一点云数据的集合所对应的待处理轨迹作为坡道区域，并将除第一点云数据之外的所有第二点云数据的集合所对应的待处理轨迹作为非坡道区域。
  13. 根据权利要求12所述的装置，其特征在于，所述区域划分单元，包括：
    索引值范围确定子单元，被配置为根据点云帧数索引值之间的连续性，将所有第一点云数据对应的点云帧数索引值划分为多个第一索引值范围，其中，第一索引值范围的个数用于表示目标车辆上坡或下坡的次数；
    坡道区域确定子单元，被配置为对于任意一个第一索引值范围，通过在该索引值范围两端位置分别增加若干个索引值，以将该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，并将各目标索引值范围所对应的待处理轨迹分别作为坡道区域；
    非坡道区域划分子单元，被配置为基于第一点云数据对应的多个目标索引值范围以及所述连续性，将除第一点云数据之外的第二点云数据对应的点云帧数索引值划分为多个第二索引值范围，并将各第二索引值范围所对应的待处理轨迹分别作为非坡道区域。
  14. 根据权利要求13所述的装置，其特征在于，所述坡道区域确定子单元，包括：
    坡道区域确定组件，被配置为对于任意一个第一索引值范围，按照该第一索引值范围中各索引值所对应的目标车辆高度信息之间的单调性，在该索引值范围两端位置分别增加符合所述单调性的索引值，以对该第一索引值范围进行扩大，得到范围扩大后的目标索引值范围，并将各目标索引值范围所对应的待处理轨迹分别作为坡道区域；其中，所述单调性包括单调递增性或单调递减性。
  15. 根据权利要求14所述的装置，其特征在于，所述坡道区域确定组件，具体被配置为：
    对于任意一个第一索引值范围，从该第一索引值范围中，确定坡道中间部位对应的中间索引值；
    确定所述中间索引值对应的目标车辆高度信息与该第一索引值范围中其他索引值对应的目标车辆高度信息之间的各高度差绝对值；
    根据车辆在坡道行驶过程中各高度差绝对值之间的单调递增性，确定在该第一索引值范围两端位置待增加的索引值；
    按照索引值的大小关系，将增加的索引值与原第一索引值范围中的索引值进行组合，以形成范围扩大后的目标索引值范围，并将各目标索引值范围所对应的待处理轨迹分别作为坡道区域。
  16. 根据权利要求10-15任一所述的装置，其特征在于，所述分层模块，包括：
    子区域确定单元，被配置为从所述非坡道区域中确定平均高度信息相同的子区域，其中，每个子区域对应的点云帧数索引值连续；
    分层单元，被配置为按照不同子区域的平均高度信息从大到小的顺序，将所述非坡道区域分为多层。
  17. 根据权利要求10所述的装置，其特征在于，所述俯仰角确定模块，具体被配置：
    基于车体坐标系与标准坐标系之间的相对关系，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角，其中，所述车体坐标系为与车体固连的坐标系，所述标准坐标系为水平面对应的坐标系；或者，
    基于惯性测量单元IMU采集的数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角；或者，
    基于多传感器融合后的数据，确定目标车辆在采集各帧点云数据时相对于水平面的车辆俯仰角，其中，所述多传感器包括IMU、GPS、雷达和/或图像传感器。
  18. 一种电子设备，其特征在于，所述电子设备包括：
    一个或多个处理器；
    存储装置，用于存储一个或多个程序，
    当所述一个或多个程序被所述一个或多个处理器执行，使得所述一个或多个处理器实现如权利要求1-9中任一所述的方法。
  19. 一种存储介质，其上存储有计算机程序，其特征在于，所述程序被处理器执行时实现如权利要求1-9中任一所述的方法。
  20. 一种车辆，其特征在于，所述车辆包含如权利要求10-17中任一所述的装置，或者包含如权利要求18所述的电子设备。
PCT/CN2022/137936 2022-04-28 2022-12-09 一种点云数据的分层方法、装置、设备、介质及车辆 Ceased WO2023207123A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP22908817.4A EP4290467A4 (en) 2022-04-28 2022-12-09 METHOD AND APPARATUS FOR OVERLAPPING POINT CLOUD DATA AND DEVICE, SUPPORT AND VEHICLE
US18/340,990 US20230331238A1 (en) 2022-04-28 2023-06-26 Layering method and apparatus for point cloud data, device, medium and vehicle

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210455299.4 2022-04-28
CN202210455299.4A CN117011447A (zh) 2022-04-28 2022-04-28 一种点云数据的分层方法、装置、设备、介质及车辆

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US18/340,990 Continuation US20230331238A1 (en) 2022-04-28 2023-06-26 Layering method and apparatus for point cloud data, device, medium and vehicle

Publications (1)

Publication Number Publication Date
WO2023207123A1 true WO2023207123A1 (zh) 2023-11-02

Family

ID=88517212

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/137936 Ceased WO2023207123A1 (zh) 2022-04-28 2022-12-09 一种点云数据的分层方法、装置、设备、介质及车辆

Country Status (2)

Country Link
CN (1) CN117011447A (zh)
WO (1) WO2023207123A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN119478281A (zh) * 2025-01-16 2025-02-18 中国铁路设计集团有限公司 一种陡峭地形下地质灾害解译方法、装置、设备及介质
CN120760728A (zh) * 2025-08-20 2025-10-10 深圳市位置科技有限公司 一种基于多传感器融合的定位方法及系统

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117760436A (zh) * 2023-12-14 2024-03-26 安徽蔚来智驾科技有限公司 车辆轨迹确定方法、存储介质及车辆

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109685898A (zh) * 2018-12-25 2019-04-26 广州文远知行科技有限公司 点云数据的分层方法、装置、计算机设备和存储介质
US20210158547A1 (en) * 2019-11-22 2021-05-27 Baidu Usa Llc Coordinate gradient method for point cloud registration for autonomous vehicles
CN112991162A (zh) * 2019-12-12 2021-06-18 北京初速度科技有限公司 一种点云数据的分层方法及装置
CN114295119A (zh) * 2021-12-31 2022-04-08 北京三快在线科技有限公司 一种构建地图的方法及装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109685898A (zh) * 2018-12-25 2019-04-26 广州文远知行科技有限公司 点云数据的分层方法、装置、计算机设备和存储介质
US20210158547A1 (en) * 2019-11-22 2021-05-27 Baidu Usa Llc Coordinate gradient method for point cloud registration for autonomous vehicles
CN112991162A (zh) * 2019-12-12 2021-06-18 北京初速度科技有限公司 一种点云数据的分层方法及装置
CN114295119A (zh) * 2021-12-31 2022-04-08 北京三快在线科技有限公司 一种构建地图的方法及装置

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN119478281A (zh) * 2025-01-16 2025-02-18 中国铁路设计集团有限公司 一种陡峭地形下地质灾害解译方法、装置、设备及介质
CN120760728A (zh) * 2025-08-20 2025-10-10 深圳市位置科技有限公司 一种基于多传感器融合的定位方法及系统

Also Published As

Publication number Publication date
CN117011447A (zh) 2023-11-07

Similar Documents

Publication Publication Date Title
US11151394B2 (en) Identifying dynamic objects in a point cloud
WO2023207123A1 (zh) 一种点云数据的分层方法、装置、设备、介质及车辆
US10545029B2 (en) Lane network construction using high definition maps for autonomous vehicles
US10846511B2 (en) Automatic detection and positioning of pole-like objects in 3D
CN112154446B (zh) 立体车道线确定方法、装置和电子设备
JP6595182B2 (ja) マッピング、位置特定、及び姿勢補正のためのシステム及び方法
EP3569983B1 (en) Algorithm and architecture for map-matching streaming probe data
US11093759B2 (en) Automatic identification of roadside objects for localization
CN110715671B (zh) 三维地图生成方法、装置、车辆导航设备和无人驾驶车辆
US12442925B2 (en) Determining weights of points of a point cloud based on geometric features
CN108981730A (zh) 用于为操作自动驾驶车辆生成参考路径的方法和系统
CN110440801A (zh) 一种定位感知信息获取方法、装置及系统
JP2020516880A (ja) ポリゴンにおける中間点を低減する方法および装置
US20190311077A1 (en) System and method for generating vehicle travel data
CN115116024B (zh) 用于车道线标注的验证方法及相关生成方法、相关设备
CN116403415B (zh) 一种车道级路况获取方法、装置、电子设备及存储介质
CN110186428A (zh) 一种基于大数据曲线拟合的道路坡度确定方法及系统
CN112446915A (zh) 一种基于图像组的建图方法及装置
CN115164918B (zh) 语义点云地图构建方法、装置及电子设备
CN117710590A (zh) 点云数据的参数化及地图构建方法
CN116147605A (zh) 一种车辆自动驾驶地图生成方法、装置、设备及存储介质
US20220228873A1 (en) Curvature value detection and evaluation
Magrath et al. Evaluating different methods for determining road grade best suited to advanced bus transportation systems
CN112163059A (zh) 基于量产车的地图数据更新方法、装置和系统
CN112530270B (zh) 一种基于区域分配的建图方法及装置

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2022908817

Country of ref document: EP

Effective date: 20230628

NENP Non-entry into the national phase

Ref country code: DE