CN116299542A - High-precision map construction method, device, storage medium and equipment - Google Patents

Abstract

The application discloses a method, a device, a storage medium and equipment for constructing a high-precision map, and belongs to the technical field of image processing. The method comprises the following steps: acquiring a point cloud set of laser radar data of a current frame and a previous frame, wherein the point cloud set comprises ground points, corner points and face points; acquiring a point cloud distance model, and calculating the distance between the same type of point clouds in the current frame and the previous frame according to the point cloud distance model; creating an error function according to the distance, and calculating the pose of the current frame relative to the previous frame according to the error function; and creating a high-precision map according to the pose. The method and the device can not only increase the diversity of the features, but also increase the constraint of the z axis through independent processing of the ground points, relieve the problem of too fast z axis divergence, and can also decouple the estimation of the pose with 6 degrees of freedom, realize the matching success rate and precision under larger accumulated error, and shorten the calculation time.

Description

High-precision map construction method, device, storage medium and equipment

technical field

The present application relates to the technical field of image processing, and in particular to a method, device, storage medium and equipment for constructing a high-precision map.

Background technique

In the automatic driving solution relying on high-precision maps, high-precision maps provide accurate prior environmental information for positioning, perception, and PnC (Planning and Control, planning and control), which have high precision, multiple elements, and high "freshness" Advantages are an important prerequisite for vehicles to make safe decisions and judgments, and are also an important guarantee for unmanned vehicles to complete various tasks safely and efficiently. Therefore, building high-precision maps is an important foundation and an important link for autonomous driving solutions.

There are generally two methods for constructing high-precision maps: one is to create high-precision maps through traditional surveying and mapping methods, specifically to collect the pose of the vehicle through high-precision inertial navigation equipment, and project the vehicle-mounted lidar data to generate High-resolution maps. The other is to create a high-precision map through SLAM (Simultaneous Localization and Mapping, simultaneous positioning and map construction) technology. Specifically, the registration method is used to estimate the pose corresponding to each frame of lidar data on the vehicle, and the loop detection and global optimization Eliminate the error and finally generate a high-precision map.

Although the traditional surveying and mapping method can quickly measure the pose of the vehicle, it is susceptible to environmental interference, especially between buildings and under dense trees. The error of the pose is often too large to meet the needs of high-precision maps. . In the case where the pose cannot be collected by high-precision inertial navigation equipment, we can use SLAM technology to accurately estimate the pose. Among them, SLAM technology includes front-end odometer. The front-end odometer can use ICP (Iterative Closest Point, iterative closest point), GICP (Generalized Iterative Closest Point, generalized iterative closest point), NDT (Normal Distribution Transform, normal distribution transformation) and other methods, but the calculation amount is relatively large, And as the scope of the map increases, the calculation efficiency is greatly reduced; therefore, the feature-based method has gradually become the mainstream method of the front-end odometer, such as LOAM (Lidar Odometry and Mapping, laser odometer and mapping), but it only extracts Corner points and surface points in the environment are used as features to estimate the pose. There are fewer constraints on the z-axis, and the cumulative error will increase with the distance moved, resulting in rapid divergence on the z-axis.

Contents of the invention

The present application provides a high-precision map construction method, device, storage medium and equipment, which are used to solve the problem of excessive z-axis divergence in the process of high-precision map construction. Described technical scheme is as follows:

On the one hand, a method for constructing a high-precision map is provided, and the method includes:

Obtain the point cloud collection of the current frame and the previous frame lidar data, the point cloud collection includes ground points, corner points and surface points;

Obtain a point cloud distance model, and calculate the distance between the current frame and the similar point cloud in the previous frame according to the point cloud distance model;

creating an error function according to the distance, and calculating the pose of the current frame relative to the previous frame according to the error function;

Create a high-resolution map from the pose.

In a possible implementation manner, the acquiring a point cloud distance model, and calculating the distance between the current frame and the same type of point cloud in the previous frame according to the point cloud distance model includes:

When the point cloud distance model is a ground point distance model, calculate the plane distance from each ground point in the point cloud set of the current frame to the ground in the previous frame, and the ground is determined by the previous frame The ground points in the point cloud collection of .

In a possible implementation manner, the acquiring a point cloud distance model, and calculating the distance between the current frame and the same type of point cloud in the previous frame according to the point cloud distance model includes:

When the point cloud distance model is a corner point distance model, calculate the angular line distance from each corner point in the point cloud set of the current frame to the corner point in the previous frame.

In a possible implementation manner, the acquiring a point cloud distance model, and calculating the distance between the current frame and the same type of point cloud in the previous frame according to the point cloud distance model includes:

When the point cloud distance model is a surface point distance model, calculate the plane distance from each surface point in the point cloud collection of the current frame to the plane in the previous frame, and the plane is determined by the previous frame The surface points in the point cloud collection of .

In a possible implementation, the method further includes:

Perform loopback detection on the point cloud collection of each frame of lidar data;

If it is detected that there is a loop between the m-th point cloud set and the n-th point cloud set, then the m-th point cloud set is matched with the ground points in the n-th point cloud set to obtain the m-th point cloud set Transformation values of the z-axis, roll angle, and pitch angle between the frame and the nth frame of lidar data;

Using the transformation values of z-axis, roll angle and pitch angle as initial values, calculate the transformation values of x-axis, y-axis and yaw angle between the mth frame and the nth frame of lidar data;

The transformation values of x-axis, y-axis, z-axis, yaw angle, roll angle and pitch angle are taken as initial values, and the final transformation value between the mth frame and the nth frame of lidar data is calculated using the ICP algorithm.

In a possible implementation, the transformation values of the z-axis, roll angle, and pitch angle are used as initial values to calculate the x-axis, y-axis, and yaw between the mth frame and the nth frame of lidar data Transformation values for angles, including:

Using the transformation values of the z-axis, roll angle and pitch angle as initial values, compressing the corner points and surface points in the m point cloud collection and the n point cloud collection into a two-dimensional map;

The two-dimensional map is matched using a CSM scanning matching algorithm to obtain transformation values of the x-axis, y-axis and yaw angle between the mth frame and the nth frame of lidar data.

In a possible implementation, the creating a high-precision map according to the pose includes:

The high-precision map is created based on the pose and the final transformation value.

On the one hand, a device for constructing a high-precision map is provided, and the device includes:

The obtaining module is used to obtain the point cloud collection of the current frame and the previous frame lidar data, and the point cloud collection includes ground points, corner points and surface points;

The obtaining module is also used to obtain a point cloud distance model, and calculate the distance between the current frame and the similar point cloud in the previous frame according to the point cloud distance model;

A calculation module, configured to create an error function according to the distance, and calculate the pose of the current frame relative to the previous frame according to the error function;

Create a module for creating a high-resolution map from the pose.

In one aspect, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, and the at least one instruction is loaded and executed by a processor to implement the method for constructing a high-precision map as described above.

In one aspect, a computer device is provided, the computer device includes a processor and a memory, at least one instruction is stored in the memory, and the instruction is loaded and executed by the processor to realize the high-precision map as described above The construction method.

The beneficial effects of the technical solution provided by the application at least include:

By dividing the point cloud in the point cloud collection into three categories: ground point, corner point and surface point, and then creating a high-precision map, this can not only increase the diversity of features, but also through the individual ground points Processing to increase the constraints of the z-axis to alleviate the problem of too fast divergence of the z-axis.

By introducing the ground point into the loop-back matching process, the estimation of the 6-DOF pose can be decoupled, the matching success rate and accuracy can be achieved under a large cumulative error, and the calculation time can be shortened.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings that need to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those skilled in the art, other drawings can also be obtained based on these drawings without creative effort.

Fig. 1 is a method flowchart of a method for constructing a high-precision map provided by an embodiment of the present application;

Fig. 2 is a method flowchart of a method for constructing a high-precision map provided by an embodiment of the present application;

Fig. 3 is a structural block diagram of a high-precision map construction device provided by another embodiment of the present application;

Fig. 4 is a structural block diagram of an apparatus for constructing a high-precision map provided by yet another embodiment of the present application.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the following will further describe the embodiments of the present application in detail in conjunction with the accompanying drawings.

Please refer to FIG. 1 , which shows a flow chart of a method for constructing a high-precision map provided by an embodiment of the present application. The method for constructing a high-precision map can be applied to a computer device. The method for constructing the high-precision map may include:

In step 101, a point cloud set of the current frame and the previous frame of lidar data is acquired, and the point cloud set includes ground points, corner points, and surface points.

The self-driving vehicle is equipped with lidar, which can periodically collect the surrounding environment, obtain lidar data, and generate a point cloud set based on each frame of lidar data.

After the computer device obtains the point cloud set, it preprocesses the point cloud, and divides the point cloud into ground points (ground) and non-ground points (non-ground). There are many methods for extracting ground points from the point cloud, which are not limited in this embodiment. For non-ground points, the computer device can divide the non-ground points into corner points and surface points according to the curvature. Therefore, a point cloud set includes ground points, corner points and surface points.

Step 102, obtain the point cloud distance model, and calculate the distance between the current frame and the same type of point cloud in the previous frame according to the point cloud distance model.

The computer device can perform modeling according to the characteristics of each category of point clouds to obtain a point cloud distance model, and use the point cloud distance model to calculate the distance between the current frame and the same type of point cloud in the previous frame. In this embodiment, the point cloud distance model created for ground points is called ground point distance model, the point cloud distance model created for corner points is called corner point distance model, and the point cloud distance model created for surface points is called is the surface point distance model. The calculation methods of the three point cloud distance models are described below.

(1) The point cloud distance model is the ground point distance model;

Specifically, obtain the point cloud distance model, and calculate the distance between the current frame and the same point cloud in the previous frame according to the point cloud distance model, which may include: when the point cloud distance model is a ground point distance model, calculate the point of the current frame The planar distance from each ground point in the cloud set to the ground in the previous frame, which is formed by the ground points in the point cloud set in the previous frame. Among them, the ground point distance model is used to estimate the plane distance between the current ground point and the ground points in the target point cloud set.

Suppose the current frame is the k+1th frame, and its corresponding point cloud set is X _k+1 , the previous frame is the kth frame, and its corresponding point cloud set is X _k , for ground point i, where i∈X _{k +1} , through the nearest neighbor search for three ground points j, l and p that are close to X _k , and these three ground points j, l and p can form a ground, therefore, the plane from the current ground point i to the ground distance

(2) The point cloud distance model is a corner point distance model;

Specifically, obtain the point cloud distance model, and calculate the distance between the current frame and the same point cloud in the previous frame according to the point cloud distance model, which may include: when the point cloud distance model is a corner point distance model, calculate the point of the current frame The angular distance from each corner point in the cloud collection to the corner point in the previous frame. Among them, the corner distance model is used to estimate the corner distance from the current corner to the target point cloud.

Suppose the current frame is the k+1th frame, and its corresponding point cloud set is X _k+1 , the previous frame is the kth frame, and its corresponding point cloud set is X _k , for corner i, where i∈X _{k +1} , through the nearest neighbor to search for two similar corner points j and l in X _k , and these two corner points j and l can form a straight line, therefore, the angular distance from the current corner point i to the connecting line

(3) The point cloud distance model is a surface point distance model;

Specifically, obtain the point cloud distance model, and calculate the distance between the current frame and the same point cloud in the previous frame according to the point cloud distance model, which may include: when the point cloud distance model is a surface point distance model, calculate the point of the current frame The plane distance from each surface point in the cloud collection to the plane in the previous frame, and the plane is composed of the surface points in the point cloud collection in the previous frame. Among them, the surface point distance model is used to estimate the plane distance between the current surface point and the surface points in the target point cloud set.

Suppose the current frame is the k+1th frame, and its corresponding point cloud set is X _k+1 , the previous frame is the kth frame, and its corresponding point cloud set is X _k , for surface point i, where i∈X _{k +1} , through the nearest neighbor search out three similar surface points j, l and p in X _k , and these three surface points j, l and p can form a plane, therefore, the plane from the current surface point i to the plane distance

Step 103, create an error function according to the distance, and calculate the pose of the current frame relative to the previous frame according to the error function.

Calculating the distance between the current frame and the same point cloud in the previous frame can also be called motion estimation, the purpose of which is to calculate the transformed pose.

The computer equipment constructs the transformation relationship f(*) for the point cloud of different features, so that:

f _c (X _{(k+1, i)} , T _k+1 )=d _c , i∈ε _k+1

f _g (X _{(k+1, i)} , T _k+1 )=d _g , i∈δ _k+1

表示当前帧的点云集合中的面点集合，δ_k+1表示当前帧的点云集合中的地面点集合。Among them, ε _k+1 represents the set of corner points in the point cloud set of the current frame,

Represents the set of surface points in the point cloud set of the current frame, and δ _k+1 represents the set of ground points in the point cloud set of the current frame.

The computer equipment can construct the error function f(T _k+1 )=d, and minimize d to 0 through the LM (Levenberg-Marquardt) method to solve the optimal pose transformation

Step 104, create a high-precision map according to the pose.

After obtaining the pose, the computer equipment can eliminate errors through loop closure detection and global optimization, and finally generate a high-precision map.

In summary, the high-precision map construction method provided by the embodiment of the present application divides the point cloud in the point cloud collection into three categories: ground points, corner points, and surface points, and then creates a high-precision map. In this way, the diversity of features can be increased, and the constraint of the z-axis can be increased by separately processing the ground points, so as to alleviate the problem of too fast divergence of the z-axis.

In addition to front-end odometer, SLAM technology also includes back-end optimization. Back-end optimization often eliminates accumulated errors by adding loopbacks. Matching methods such as ICP are often used for loopbacks. However, if the cumulative error between two frames of lidar data is too large, it is difficult to calculate the correct loopback pose. For this reason, we can also use a global matching method to estimate the pose, such as Cartographer, which uses CSM (Correlation Scan Match, Correlation Scan Match) for pose estimation, but it takes a lot of time to estimate the 6 degrees of freedom increases, reducing the overall computational efficiency of mapping. In order to solve the problem of excessive consumption of loopback matching resources, this embodiment provides a loopback detection method, please refer to Figure 2 for the specific process:

Step 201, performing loop closure detection on the point cloud set of each frame of lidar data.

The computer device can perform loopback detection on the point cloud set through many methods, which are not limited in this embodiment.

If a loop is detected, the computer device will obtain a point cloud set of two frames of lidar data, and the transformation between the two point cloud sets needs to be calculated below.

Step 202, if it is detected that there is a loop between the m-th point cloud set and the n-th point cloud set, then match the ground points in the m-th point cloud set and the n-th point cloud set to obtain the m-th frame The transformation values of the z-axis, roll angle, and pitch angle between the n-th frame of lidar data.

Assuming that the detected two frames of lidar data are the m-th frame and the n-th frame of lidar data, the corresponding points are the m-th and n-th point cloud sets. The computer device can fix one point cloud set as the target point cloud set, and calculate the transformation of another point cloud set relative to the target point cloud set. Among them, the calculation process is divided into two parts, the first part is to match the ground points in the mth and nth point cloud sets, and the second part is to match the corner points in the mth and nth point cloud sets Match with pasta.

When solving the first part, the computer equipment uses the Normal-Icp method to process the ground points, and preferentially estimates the transformation values of the three degrees of freedom [z, roll, pitch]. Among them, z represents the z-axis, roll represents the roll angle, and pitch represents the pitch angle.

Step 203, using the transformation values of the z-axis, roll angle and pitch angle as initial values, and calculating the transformation values of the x-axis, y-axis and yaw angle between the mth frame and the nth frame of lidar data.

Specifically, the computer device can use the transformation values of the z-axis, roll angle, and pitch angle as initial values, and compress the corner points and surface points in the m-th point cloud set and the n-th point cloud set into a two-dimensional map; The CSM scan matching algorithm is used to match the two-dimensional map, and the transformation values of the x-axis, y-axis and yaw angle between the m-th frame and the n-th frame of lidar data are obtained.

Wherein, the computer device can fix the transformation values of the z-axis, roll angle and pitch angle as initial values, and further process the corner points and surface points. After compressing the corner points and surface points into a two-dimensional map, the computer device can use the CSM method to perform matching to calculate the transformation values of the three degrees of freedom [x, y, yaw]. Among them, x represents the x-axis, y represents the y-axis, and yaw represents the yaw angle.

Step 204, using the transformation values of the x-axis, y-axis, z-axis, yaw angle, roll angle and pitch angle as initial values, using the ICP algorithm to calculate the final transformation value between the mth frame and the nth frame of lidar data .

The computer equipment combines the transformation values obtained from the two calculations to finally obtain the transformation values of the 6 degrees of freedom [x, y, z, roll, pitch, yaw]. Since there will be discretization errors in the calculation process, the computer equipment also needs to use the transformation value of 6 degrees of freedom as the initial value, and then perform an ICP match on the mth and nth frames of lidar data to obtain the final The exact two-frame transformation of lidar data. On the one hand, this loopback matching method can improve the matching success rate and accuracy under a large cumulative error; on the other hand, it also greatly shortens the calculation time and improves the efficiency.

If loop-closing matching is introduced, step 104 can be replaced by: creating a high-precision map according to the pose and the final transformation value.

In this embodiment, by introducing the ground point into the loop-back matching link, the estimation of the 6-DOF pose can be decoupled, the matching success rate and accuracy can be achieved under a large cumulative error, and the calculation time can be shortened.

Please refer to FIG. 3 , which shows a structural block diagram of an apparatus for constructing a high-precision map provided by an embodiment of the present application. The apparatus for constructing a high-precision map can be applied to a computer device. The device for constructing the high-precision map may include:

Obtaining module 310, used to obtain the point cloud collection of the current frame and the previous frame lidar data, the point cloud collection includes ground points, corner points and surface points;

The obtaining module 310 is also used to obtain the point cloud distance model, and calculates the distance between the current frame and the previous frame according to the point cloud distance model;

Calculation module 320, is used for creating error function according to distance, calculates the pose of current frame relative to previous frame according to error function;

The creation module 330 is used to create a high-precision map according to the pose.

In an optional embodiment, the calculating module 320 is also used for:

When the point cloud distance model is the ground point distance model, calculate the plane distance from each ground point in the point cloud set of the current frame to the ground in the previous frame, and the ground is composed of the ground points in the point cloud set of the previous frame.

In an optional embodiment, the calculating module 320 is also used for:

When the point cloud distance model is a corner point distance model, calculate the angular line distance from each corner point in the point cloud collection of the current frame to the corner point of the previous frame.

In an optional embodiment, the calculating module 320 is also used for:

When the point cloud distance model is the surface point distance model, calculate the plane distance from each surface point in the point cloud collection of the current frame to the plane in the previous frame, and the plane is formed by the surface points in the point cloud collection of the previous frame.

Please refer to Figure 4, in an optional embodiment, the device also includes:

The detection module 340 is used to perform loopback detection on the point cloud collection of each frame of lidar data;

The matching module 350 is used to match the ground points in the m point cloud set and the n point cloud set if it is detected that there is a loop between the m point cloud set and the n point cloud set to obtain Transformation values of the z-axis, roll angle, and pitch angle between the mth frame and the nth frame of lidar data;

The calculation module 320 is also used to use the transformation values of the z-axis, roll angle and pitch angle as initial values, and calculate the transformation values of the x-axis, y-axis and yaw angle between the mth frame and the nth frame of lidar data ;

The calculation module 320 is also used to use the transformation values of the x-axis, y-axis, z-axis, yaw angle, roll angle and pitch angle as initial values, and use the ICP algorithm to calculate the distance between the mth frame and the nth frame of lidar data. The final transformation value of .

In an optional embodiment, the calculating module 320 is also used for:

Using the transformation values of the z-axis, roll angle, and pitch angle as initial values, compress the corner points and surface points in the m-th point cloud collection and the n-th point cloud collection into a two-dimensional map;

The CSM scan matching algorithm is used to match the two-dimensional map, and the transformation values of the x-axis, y-axis and yaw angle between the m-th frame and the n-th frame of lidar data are obtained.

In an optional embodiment, the creation module 330 is also used for:

Create a high-resolution map from the pose and final transformation values.

In summary, the high-precision map construction device provided by the embodiment of the present application divides the point cloud in the point cloud collection into three categories: ground points, corner points, and surface points, and then creates a high-precision map. In this way, the diversity of features can be increased, and the constraint of the z-axis can be increased by separately processing the ground points, so as to alleviate the problem of too fast divergence of the z-axis.

By introducing the ground point into the loop-back matching process, the estimation of the 6-DOF pose can be decoupled, the matching success rate and accuracy can be achieved under a large cumulative error, and the calculation time can be shortened.

An embodiment of the present application provides a computer-readable storage medium, at least one instruction is stored in the storage medium, and the at least one instruction is loaded and executed by a processor to implement the above-mentioned method for constructing a high-precision map.

An embodiment of the present application provides a computer device, the computer device includes a processor and a memory, at least one instruction is stored in the memory, and the instruction is loaded and executed by the processor to realize the above-mentioned high-level The construction method of the precision map.

It should be noted that: when the device for constructing a high-precision map provided by the above-mentioned embodiment constructs a high-precision map, the division of the above-mentioned functional modules is used as an example for illustration. In practical applications, the above-mentioned functions can be allocated by Different functional modules are completed, that is, the internal structure of the high-precision map construction device is divided into different functional modules to complete all or part of the functions described above. In addition, the high-precision map construction device and the high-precision map construction method embodiment provided by the above-mentioned embodiments belong to the same concept, and its specific implementation process is detailed in the method embodiment, and will not be repeated here.

Those of ordinary skill in the art can understand that all or part of the steps for implementing the above embodiments can be completed by hardware, and can also be completed by instructing related hardware through a program. The program can be stored in a computer-readable storage medium. The above-mentioned The storage medium mentioned may be a read-only memory, a magnetic disk or an optical disk, and the like.

The above description is not intended to limit the embodiments of the present application, and any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the embodiments of the present application shall be included within the scope of protection of the embodiments of the present application.

Claims (10)

1. The method for constructing the high-precision map is characterized by comprising the following steps of:

acquiring a point cloud set of laser radar data of a current frame and a previous frame, wherein the point cloud set comprises ground points, corner points and face points;

acquiring a point cloud distance model, and calculating the distance between the same type of point clouds in the current frame and the previous frame according to the point cloud distance model;

creating an error function according to the distance, and calculating the pose of the current frame relative to the previous frame according to the error function;

and creating a high-precision map according to the pose.

2. The method for constructing a high-precision map according to claim 1, wherein the obtaining a point cloud distance model, and calculating the distance between the current frame and the similar point clouds in the previous frame according to the point cloud distance model, comprises:

and when the point cloud distance model is a ground point distance model, calculating the plane distance between each ground point in the point cloud set of the current frame and the ground in the previous frame, wherein the ground is formed by the ground points in the point cloud set of the previous frame.

3. The method for constructing a high-precision map according to claim 1, wherein the obtaining a point cloud distance model, and calculating the distance between the current frame and the similar point clouds in the previous frame according to the point cloud distance model, comprises:

and when the point cloud distance model is an angular point distance model, calculating the angular line distance from each angular point in the point cloud set of the current frame to the angular point of the previous frame.

4. The method for constructing a high-precision map according to claim 1, wherein the obtaining a point cloud distance model, and calculating the distance between the current frame and the similar point clouds in the previous frame according to the point cloud distance model, comprises:

and when the point cloud distance model is a face point distance model, calculating the plane distance between each face point in the point cloud set of the current frame and the plane in the previous frame, wherein the plane is formed by the face points in the point cloud set of the previous frame.

5. The method of constructing a high-precision map according to any one of claims 1 to 4, further comprising:

performing loop detection on the point cloud set of each frame of laser radar data;

if a loop exists between the mth point cloud set and the nth point cloud set, matching ground points in the mth point cloud set and the nth point cloud set to obtain a conversion value of a z-axis, a roll angle and a pitch angle between mth frame and nth frame laser radar data;

taking the conversion values of the z-axis, the roll angle and the pitch angle as initial values, and calculating the conversion values of the x-axis, the y-axis and the yaw angle between the laser radar data of the mth frame and the laser radar data of the nth frame;

and calculating a final conversion value between the m-th frame and the n-th frame of laser radar data by using the conversion values of the x-axis, the y-axis, the z-axis, the yaw angle, the roll angle and the pitch angle as initial values through an ICP algorithm.

6. The method according to claim 5, wherein calculating the conversion values of the x-axis, y-axis and yaw angle between the mth frame and the nth frame of lidar data using the conversion values of the z-axis, roll angle and pitch angle as initial values comprises:

the transformation values of the z-axis, the roll angle and the pitch angle are used as initial values, and the corner points and the face points in the mth point cloud set and the nth point cloud set are compressed into a two-dimensional map;

and matching the two-dimensional map by using a CSM scanning matching algorithm to obtain conversion values of an x axis, a y axis and a yaw angle between the laser radar data of the mth frame and the laser radar data of the nth frame.

7. The method of claim 5, wherein creating the high-precision map from the pose comprises:

and creating the high-precision map according to the pose and the final transformation value.

8. A high-precision map construction apparatus, characterized in that the apparatus comprises:

the acquisition module is used for acquiring a point cloud set of the laser radar data of the current frame and the previous frame, wherein the point cloud set comprises ground points, corner points and face points;

the acquisition module is further used for acquiring a point cloud distance model, and calculating the distance between the current frame and the similar point cloud in the previous frame according to the point cloud distance model;

the calculating module is used for creating an error function according to the distance and calculating the pose of the current frame relative to the previous frame according to the error function;

and the creation module is used for creating a high-precision map according to the pose.

9. A computer readable storage medium having stored therein at least one instruction that is loaded and executed by a processor to implement the method of constructing a high precision map according to any one of claims 1 to 7.

10. A computer device comprising a processor and a memory, the memory having stored therein at least one instruction that is loaded and executed by the processor to implement the method of constructing a high precision map according to any one of claims 1 to 7.

Images