Point cloud map processing method and device, equipment and storage medium

By performing type detection and feature extraction on the sub-trajectories of vehicle motion trajectories, combined with point cloud distribution processing and geometric feature extraction, the problem of insufficient generation speed and accuracy of point cloud maps in the existing technology is solved, and the generation of point cloud maps is improved Effect.

CN118941598APending Publication Date: 2024-11-12TENCENT TECHNOLOGY (SHENZHEN) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310531577.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-10
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing point cloud SLAM algorithms lack processing capabilities in some scenarios or feature extraction, resulting in low point cloud map generation speed and accuracy.

Method used

By performing type detection of the subtrajectories of the vehicle motion trajectory, point cloud distribution processing and geometric feature extraction are performed on the subtrajectories of the target type, the first point cloud feature and the second point cloud feature are obtained, and used to match the point cloud map The registration improves the generation speed and accuracy of point cloud maps.

Benefits of technology

It improves the generation speed and accuracy of point cloud maps, enhances the feature extraction capability of point cloud maps, and improves the real-time performance of the system and the reasonable allocation of computing resources.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118941598A_ABST
    Figure CN118941598A_ABST
Patent Text Reader

Abstract

The invention relates to a point cloud map processing method and device, computer equipment, a storage medium and a computer program product, and the method can be applied to the field of maps, and comprises the steps: obtaining a motion track of a vehicle and a point cloud data set collected according to the motion track; performing type detection on the sub-tracks of the motion track to obtain target sub-tracks belonging to a target type; performing point cloud distribution processing on to-be-processed point cloud data in the point cloud data set to obtain a first point cloud feature of the to-be-processed point cloud data; when the sub-track corresponding to the to-be-processed point cloud data is the target sub-track, performing geometric feature extraction on the to-be-processed point cloud data to obtain a second point cloud feature of the to-be-processed point cloud data; and registering the to-be-processed point cloud data with a point cloud map based on the first point cloud feature and the second point cloud feature to obtain an updated point cloud map. By adopting the method, the generation speed and accuracy of the point cloud map can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, device, equipment and storage medium for processing a point cloud map. Background Art

[0002] A high-precision map is a map with higher spatial accuracy and richer information than traditional maps. It provides a detailed three-dimensional description of geographical elements such as roads, traffic signs, road signs, buildings, etc., and is therefore widely used in the fields of autonomous driving, robot navigation, and drone flight. As a form of high-precision map, point cloud maps can provide rich three-dimensional information to help the system achieve accurate positioning and perception.

[0003] At present, point cloud SLAM (Simultaneous Localization and Mapping) algorithms are usually used to perform real-time positioning and build point cloud maps based on the collected point cloud data. However, existing point cloud SLAM algorithms usually have insufficient processing capabilities in certain scenes or feature extraction, resulting in low speed and accuracy in generating point cloud maps. Summary of the invention

[0004] Based on this, it is necessary to provide a point cloud map processing method, device, equipment, storage medium and computer program product that can improve the point cloud map generation effect in response to the above technical problems.

[0005] In a first aspect, the present application provides a method for processing a point cloud map. The method comprises:

[0006] Acquire a motion trajectory of the vehicle and a point cloud data set collected according to the motion trajectory;

[0007] Performing type detection on the sub-trajectories of the motion trajectory to obtain target sub-trajectories belonging to the target type;

[0008] Performing point cloud distribution processing on the point cloud data to be processed in the point cloud data set to obtain a first point cloud feature of the point cloud data to be processed;

[0009] When the sub-track corresponding to the point cloud data to be processed is the target sub-track, geometric feature extraction is performed on the point cloud data to be processed to obtain a second point cloud feature of the point cloud data to be processed;

[0010] The point cloud data to be processed is registered with a point cloud map based on the first point cloud feature and the second point cloud feature to obtain an updated point cloud map.

[0011] In a second aspect, the present application also provides a point cloud map processing device. The device comprises:

[0012] A data acquisition module, used to acquire the motion trajectory of the vehicle and the point cloud data set acquired according to the motion trajectory;

[0013] A type detection module, used for performing type detection on the sub-trajectory of the motion trajectory to obtain a target sub-trajectory belonging to a target type;

[0014] A first point cloud feature extraction module is used to perform point cloud distribution processing on the point cloud data to be processed in the point cloud data set to obtain a first point cloud feature of the point cloud data to be processed;

[0015] A second point cloud feature extraction module is used to extract geometric features of the point cloud data to be processed when the sub-track corresponding to the point cloud data to be processed is the target sub-track, so as to obtain a second point cloud feature of the point cloud data to be processed;

[0016] A map updating module is used to align the point cloud data to be processed with the point cloud map based on the first point cloud feature and the second point cloud feature to obtain an updated point cloud map.

[0017] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0018] Acquire a motion trajectory of the vehicle and a point cloud data set collected according to the motion trajectory;

[0019] Performing type detection on the sub-trajectories of the motion trajectory to obtain target sub-trajectories belonging to the target type;

[0020] Performing point cloud distribution processing on the point cloud data to be processed in the point cloud data set to obtain a first point cloud feature of the point cloud data to be processed;

[0021] When the sub-track corresponding to the point cloud data to be processed is the target sub-track, geometric feature extraction is performed on the point cloud data to be processed to obtain a second point cloud feature of the point cloud data to be processed;

[0022] The point cloud data to be processed is registered with a point cloud map based on the first point cloud feature and the second point cloud feature to obtain an updated point cloud map.

[0023] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0024] Acquire a motion trajectory of the vehicle and a point cloud data set collected according to the motion trajectory;

[0025] Performing type detection on the sub-trajectories of the motion trajectory to obtain target sub-trajectories belonging to the target type;

[0026] Performing point cloud distribution processing on the point cloud data to be processed in the point cloud data set to obtain a first point cloud feature of the point cloud data to be processed;

[0027] When the sub-track corresponding to the point cloud data to be processed is the target sub-track, geometric feature extraction is performed on the point cloud data to be processed to obtain a second point cloud feature of the point cloud data to be processed;

[0028] The point cloud data to be processed is registered with a point cloud map based on the first point cloud feature and the second point cloud feature to obtain an updated point cloud map.

[0029] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0030] Acquire a motion trajectory of the vehicle and a point cloud data set collected according to the motion trajectory;

[0031] Performing type detection on the sub-trajectories of the motion trajectory to obtain target sub-trajectories belonging to the target type;

[0032] Performing point cloud distribution processing on the point cloud data to be processed in the point cloud data set to obtain a first point cloud feature of the point cloud data to be processed;

[0033] When the sub-track corresponding to the point cloud data to be processed is the target sub-track, geometric feature extraction is performed on the point cloud data to be processed to obtain a second point cloud feature of the point cloud data to be processed;

[0034] The point cloud data to be processed is registered with a point cloud map based on the first point cloud feature and the second point cloud feature to obtain an updated point cloud map.

[0035] The above-mentioned point cloud map processing method, device, computer equipment, storage medium and computer program product, after obtaining the vehicle's motion trajectory and the point cloud data set collected according to the motion trajectory, obtains a target sub-trajectory belonging to the target type (such as a sub-trajectory corresponding to a scene where the existing point cloud SLAM algorithm has insufficient processing capabilities) by performing type detection on the sub-trajectory of the motion trajectory. For the target sub-trajectory of the target type, on the one hand, point cloud distribution processing is performed on the point cloud data to be processed of the target sub-trajectory to obtain a first point cloud feature of the point cloud data to be processed; on the other hand, geometric feature extraction is performed on the point cloud data to be processed of the target sub-trajectory to obtain a second point cloud feature of the point cloud data to be processed, thereby better extracting the features of the point cloud data to be processed, and fusing the first point cloud feature and the second point cloud feature to align the point cloud data to be processed with the point cloud map, thereby improving the accuracy and stability of the alignment, and thereby improving the generation speed and accuracy of the point cloud map. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A diagram of an application environment of a method for processing a point cloud map in one embodiment;

[0037] Figure 2 A schematic diagram of a process of processing a point cloud map in one embodiment;

[0038] Figure 3 is a schematic diagram of a motion trajectory in one embodiment;

[0039] Figure 4 is a schematic diagram of a motion trajectory in another embodiment;

[0040] Figure 5 is a schematic diagram of point cloud data in one embodiment;

[0041] Figure 6 is a schematic diagram of a point cloud feature matching result in one embodiment;

[0042] Figure 7 A schematic diagram of a point cloud data stitching result in one embodiment;

[0043] Figure 8 is a flowchart of a method for processing a point cloud map in another embodiment;

[0044] Fig. 9 A schematic diagram of fusion SLAM in an embodiment;

[0045] Fig.10 A schematic diagram of a positioning trajectory test result in an embodiment;

[0046] Fig.11 A schematic diagram of positioning trajectory test results in another embodiment;

[0047] Fig.12 A schematic diagram of a point cloud stitching test result in one embodiment;

[0048] Fig.13 A schematic diagram of a point cloud stitching test result in another embodiment;

[0049] Fig.14 is a structural block diagram of a point cloud map processing device in one embodiment;

[0050] Fig.15 It is a structural block diagram of a point cloud map processing device in another embodiment;

[0051] Fig.16 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0053] The point cloud map processing method provided in this application can be applied to the field of intelligent transportation, and specifically to the field of dynamic driving of intelligent transportation.

[0054] Intelligent Traffic System (ITS), also known as Intelligent Transportation System, is an effective and comprehensive application of advanced science and technology (information technology, computer technology, data communication technology, sensor technology, electronic control technology, automatic control theory, operations research, artificial intelligence, etc.) to transportation, service control and vehicle manufacturing, strengthening the connection between vehicles, roads and users, thus forming a comprehensive transportation system that ensures safety, improves efficiency, improves the environment and saves energy. Or;

[0055] Intelligent Vehicle Infrastructure Cooperative Systems (IVICS), referred to as IVICS, is a development direction of Intelligent Transportation Systems (ITS). IVICS uses advanced wireless communications and new generation Internet technologies to implement all-round dynamic real-time information interaction between vehicles and roads, and conducts active vehicle safety control and road cooperative management based on the collection and integration of dynamic traffic information in all time and space, fully realizing the effective coordination of people, vehicles and roads, ensuring traffic safety, and improving traffic efficiency, thus forming a safe, efficient and environmentally friendly road traffic system.

[0056] Autonomous driving, also known as driverless or self-driving cars, is a technology that uses various sensors, controllers, computer vision and artificial intelligence technologies to achieve autonomous driving of vehicles. Autonomous driving cars are able to complete driving tasks without human intervention, achieving safe and efficient travel from starting point to destination.

[0057] The point cloud map processing method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other servers. The processing method of the point cloud map is executed by the terminal 102 or the server 104 alone, or by the terminal 102 and the server 104 in collaboration. In some embodiments, the processing method of the point cloud map is executed by the server 102, and the terminal 102 obtains the motion trajectory of the vehicle and the point cloud data set collected according to the motion trajectory; performs type detection on the sub-trajectory of the motion trajectory to obtain a target sub-trajectory belonging to the target type; performs point cloud distribution processing on the point cloud data to be processed in the point cloud data set to obtain the first point cloud feature of the point cloud data to be processed; when the sub-trajectory corresponding to the point cloud data to be processed is the target sub-trajectory, the geometric feature extraction of the point cloud data to be processed is performed to obtain the second point cloud feature of the point cloud data to be processed; based on the first point cloud feature and the second point cloud feature, the point cloud data to be processed is aligned with the point cloud map to obtain an updated point cloud map.

[0058] Among them, the terminal 102 can be, but is not limited to, various desktop computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The terminal 102 and the server 104 can be directly or indirectly connected via wired or wireless communications, and this application does not limit this.

[0059] In one embodiment, Figure 2 As shown, a method for processing a point cloud map is provided, and the method is applied to Figure 1 The computer device in the example is used to illustrate, including the following steps:

[0060] S202, obtaining a motion trajectory of the vehicle and a point cloud data set collected according to the motion trajectory.

[0061] The vehicle can be a collection vehicle or other vehicles, such as an autonomous driving vehicle. A collection vehicle is a vehicle specifically used to collect road environment data. It usually has a variety of sensor devices, such as lidar, camera, GPS, IMU, etc., which can obtain the vehicle's position, posture, speed and other information in real time, and obtain three-dimensional point cloud data of the road environment.

[0062] The collection vehicle usually drives on the road, scanning and collecting the road environment through multiple sensor devices. Equipment such as laser radar and camera can be used to obtain three-dimensional point cloud data and image data of the road environment, and equipment such as GPS and IMU can be used to obtain the vehicle's position and posture information, thereby constructing the vehicle's motion trajectory.

[0063] In addition to data collection vehicles, other vehicles can also obtain vehicle motion trajectories and point cloud data by installing various sensor devices. For example, autonomous vehicles can obtain road environment data through sensor devices such as lidar and cameras to achieve autonomous navigation and control of the vehicle. It should be noted that compared with data collection vehicles, other vehicles usually require more complex algorithm design and system integration to achieve accurate and stable data collection.

[0064] The motion trajectory is a continuous sequence of position points that a vehicle passes through during its driving process. Specifically, a series of position points can be used to represent the vehicle's driving route and path.

[0065] A point cloud dataset refers to a data set consisting of multiple frames of point cloud data, usually containing point cloud data collected at multiple time points, which is used to describe the road environment during vehicle driving. Each frame of point cloud data contains the three-dimensional point cloud information of the road environment obtained from the sensor at the corresponding time point.

[0066] Specifically, the terminal controls the vehicle to drive in the target environment, and during the driving process, collects the vehicle's position and posture data through the GPS and IMU devices configured on the vehicle, and then determines the vehicle's motion trajectory based on the collected vehicle's position and posture data, and collects point cloud data of the vehicle's environment through the laser radar configured on the vehicle.

[0067] like Figure 3 The figure shows a schematic diagram of a motion trajectory in an embodiment, wherein the dotted line is the motion trajectory of the vehicle.

[0068] S204, performing type detection on the sub-trajectories of the motion trajectory to obtain target sub-trajectories belonging to the target type.

[0069] Among them, a sub-trajectory refers to a continuous subsequence in a complete motion trajectory, usually a trajectory segment between two time points or a trajectory segment within a specific interval. Target types can specifically include turning type, weak texture type, low speed type and initialization type.

[0070] Specifically, after obtaining the motion trajectory of the vehicle, the terminal may divide the motion trajectory to obtain at least two sub-trajectories, and perform type detection on each of the sub-trajectories obtained by the division to obtain the type corresponding to each sub-trajectory, and determine the target sub-trajectory belonging to the target type.

[0071] refer to Figure 4 The motion trajectory diagram shown is Figure 4 The sub-trajectory corresponding to the circular box is the target sub-trajectory of the initial type, and the sub-trajectory corresponding to the elliptical box is the target sub-trajectory of the turning type.

[0072] S206, performing point cloud distribution processing on the point cloud data to be processed in the point cloud data set to obtain a first point cloud feature of the point cloud data to be processed.

[0073] It should be noted that the motion trajectory corresponds to a point cloud data set, and each sub-trajectory of the motion trajectory corresponds to a point cloud data subset. In the embodiment of the present application, the processing process of the point cloud data set can be specifically performed on a point cloud data subset basis. For example, the first point cloud data subset corresponding to the first sub-trajectory is processed first, and then the second point cloud data subset corresponding to the second sub-trajectory is processed, and so on until all point cloud data subsets of the point cloud data set are processed; for the point cloud data subset corresponding to any sub-trajectory, the process of processing each frame of point cloud data in the point cloud data subset can be performed frame by frame, for example, the first frame of point cloud data in the point cloud data subset is processed first, and then the second frame of point cloud data in the point cloud data subset is processed, ..., the i-th frame of point cloud data is processed, ..., until the w frames of point cloud data in the point cloud data subset are processed.

[0074] The point cloud data to be processed refers to the point cloud data that needs to be processed and analyzed, which can be the point cloud data of the current frame. For example, the point cloud data set corresponding to the motion trajectory contains 20 frames of point cloud data. The processing of the first 9 frames of point cloud data has been completed, and the point cloud data of the 10th frame is the point cloud data to be processed.

[0075] Point cloud distribution processing refers to the distribution analysis and processing of point cloud data to obtain feature information of point cloud data and realize applications such as registration, classification, and segmentation of point cloud data. In the embodiment of the present application, point cloud distribution processing can specifically be NDT (Normal Distribution Transform) processing. The NDT algorithm is a commonly used point cloud registration and feature extraction method, which can be used for point cloud distribution processing to obtain feature information of point cloud data.

[0076] The NDT algorithm is a point cloud registration algorithm based on the Gaussian distribution model. It mainly models the distribution of point cloud data as multiple Gaussian distributions, and realizes the registration and matching of point cloud data by matching the similarity between different Gaussian distributions. In the NDT algorithm, point cloud data can be regarded as a probability density function. By estimating parameters such as the mean and covariance between different Gaussian distributions, feature extraction and matching of point cloud data can be realized. The first point cloud feature can specifically be the spatial distribution feature of the point cloud data to be processed.

[0077] Specifically, after obtaining the point cloud data to be processed, the terminal may preprocess the point cloud data to obtain preprocessed point cloud data, and perform point cloud distribution processing on the preprocessed point cloud data, thereby obtaining the first point cloud feature of the point cloud data to be processed.

[0078] Among them, preprocessing can be specifically denoising, downsampling, conversion of coordinate system, etc. Denoising refers to denoising the point cloud data to be processed, such as using statistical filters, radius filters and other methods to eliminate noise points in the point cloud and improve the quality of the point cloud data; downsampling refers to downsampling the point cloud data, such as using a voxel grid filter to aggregate the points in the point cloud data set into voxels to reduce the amount of data, which can speed up the subsequent processing and reduce the computational complexity; conversion of coordinate system refers to converting the point cloud data to be processed from its original coordinate system to a coordinate system consistent with the point cloud map. Figure 1 This helps to align point cloud data in the subsequent feature extraction and registration process.

[0079] S208 , when the sub-trajectory corresponding to the point cloud data to be processed is the target sub-trajectory, geometric feature extraction is performed on the point cloud data to be processed to obtain a second point cloud feature of the point cloud data to be processed.

[0080] Geometric feature extraction refers to extracting features related to the geometric shape and structure of the point cloud from point cloud data. In the embodiment of the present application, geometric feature extraction can be specifically extracted by LOAM (Lidar Odometry and Mapping). LOAM is a SLAM algorithm based on laser radar, which achieves high-precision positioning and mapping by extracting geometric features of point cloud data. In the LOAM algorithm, geometric feature extraction mainly includes the extraction of edge features and plane features.

[0081] Edge features usually correspond to areas such as sharp angles and sharp edges in point cloud data. In the LOAM algorithm, the extraction of edge features mainly depends on the curvature information of the point cloud data. Specifically, LOAM calculates the curvature of each point and sorts the points according to the size of the curvature value. Points with larger curvature are considered to be edge feature points; plane features usually correspond to smooth, planar areas in point cloud data. In the LOAM algorithm, the extraction of plane features also depends on the curvature information of the point cloud data. Points with smaller curvature are considered to be plane feature points. By extracting edge features and plane features, the LOAM algorithm can extract features related to the geometric shape and structure of the point cloud from the point cloud data. These features can be used for subsequent point cloud matching, positioning and mapping tasks. The second point cloud feature can specifically be the spatial distribution feature of the point cloud data to be processed.

[0082] Specifically, when the terminal processes any frame of point cloud data, the frame of point cloud data is the point cloud data to be processed, and the sub-track corresponding to the point cloud data to be processed is determined, and the type corresponding to the sub-track is determined. If the sub-track is the target type, the point cloud data to be processed is pre-processed to obtain the pre-processed point cloud data, and then the geometric features of the pre-processed point cloud data are extracted to obtain the second point cloud features of the point cloud data to be processed.

[0083] S210: aligning the point cloud data to be processed with the point cloud map based on the first point cloud feature and the second point cloud feature to obtain an updated point cloud map.

[0084] Among them, the point cloud map refers to an existing point cloud map, which is a point cloud map obtained by updating frame by frame based on the processed point cloud data in the point cloud data set. For example, the point cloud data set corresponding to the motion trajectory contains 20 frames of point cloud data, and the processing of the first 9 frames of point cloud data has been completed. Then the point cloud map is obtained based on the first 9 frames of point cloud data.

[0085] Specifically, after obtaining the first point cloud feature and the second point cloud feature, the terminal determines the radar posture of the vehicle based on the first point cloud feature and the second point cloud feature, and updates the point cloud data to be processed to the point cloud map based on the radar posture of the vehicle to obtain an updated point cloud map.

[0086] In the above embodiment, after obtaining the motion trajectory of the vehicle and the point cloud data set collected according to the motion trajectory, the terminal performs type detection on the sub-trajectory of the motion trajectory to obtain a target sub-trajectory belonging to the target type (such as a sub-trajectory corresponding to a scene where the existing point cloud SLAM algorithm has insufficient processing capabilities). For the target sub-trajectory of the target type, on the one hand, point cloud distribution processing is performed on the point cloud data to be processed of the target sub-trajectory to obtain a first point cloud feature of the point cloud data to be processed; on the other hand, geometric feature extraction is performed on the point cloud data to be processed of the target sub-trajectory to obtain a second point cloud feature of the point cloud data to be processed, thereby better extracting the features of the point cloud data to be processed, and fusing the first point cloud feature and the second point cloud feature to align the point cloud data to be processed with the point cloud map, thereby improving the accuracy and stability of the alignment, and thereby improving the generation speed and accuracy of the point cloud map.

[0087] In one embodiment, when the sub-track corresponding to the point cloud data to be processed is not the target sub-track, the terminal may directly register the point cloud data to be processed with the point cloud map based on the first point cloud feature to obtain an updated point cloud map.

[0088] It is understandable that when the sub-trajectory corresponding to the point cloud data to be processed is not the target sub-trajectory, that is, when the sub-trajectory corresponding to the point cloud data to be processed does not have such rich geometric features in the point cloud data (such as a straight driving trajectory), or their geometric features contribute less to map construction and positioning, in this case, the point cloud data to be processed can be directly aligned with the point cloud map based on the first point cloud feature to obtain an updated point cloud map, avoiding complex geometric feature extraction on these non-target sub-trajectories, thereby reducing the computational complexity and improving the real-time performance of the system; by making a trade-off between the target sub-trajectory and the non-target sub-trajectory, it is possible to achieve a reasonable allocation of computing resources while maintaining map accuracy, thereby improving the overall system performance.

[0089] In one embodiment, the target type includes a turning type, and the terminal performs type detection on a sub-trajectory of the motion trajectory, and a process of obtaining a target sub-trajectory belonging to the target type specifically includes the following steps: determining a curvature value of each sub-trajectory in the motion trajectory; in the motion trajectory, determining a sub-trajectory corresponding to a curvature value that meets a curvature condition as a target sub-trajectory of the turning type.

[0090] The curvature value is a quantity used to describe the curvature of the sub-track. The curvature condition can specifically pre-set a curvature threshold value, which is used to determine whether the sub-track is a target sub-track of a turning type. It can be understood that when the curvature value of the sub-track is greater than or equal to the threshold value, the sub-track can be considered to meet the curvature condition, and thus it is considered to be a target sub-track of a turning type.

[0091] Specifically, the terminal determines the curvature value of each point on the sub-trajectory, and determines the curvature value of the corresponding sub-trajectory based on the curvature value of each point, and then determines whether the curvature value of the sub-trajectory meets the curvature condition, and determines the sub-trajectory corresponding to the curvature value that meets the curvature condition as the target sub-trajectory of the turning type.

[0092] Among them, for any sub-trajectory, the curvature value of the sub-trajectory is determined based on the curvature value of each point. Specifically, the average value of the curvature values ​​of each point can be determined, and the obtained average value is determined as the curvature value of the sub-trajectory, or the maximum value of the curvature values ​​of each point is determined as the curvature value of the sub-trajectory.

[0093] In one embodiment, for the i-th point on the sub-track, the curvature value corresponding to the i-th track point in the sub-track may be determined by the following formula:

[0094]

[0095] in, is the vector connecting the i-1th point to the i-th point on the sub-trajectory, is the vector connecting the i-th point to the i+1-th point on the sub-trajectory, θ is the vector and vector In the embodiment of the present application, the vector and vector The angle θ between them is determined as the curvature value of the i-th point.

[0096] It should be noted that in the motion trajectory, the turning sub-trajectory usually has a higher curvature. By calculating the curvature value and setting the curvature condition, the turning sub-trajectory can be effectively identified and distinguished. The turning sub-trajectory often has richer geometric features in the point cloud data. By distinguishing the turning sub-trajectory and extracting the first point cloud feature and the second point cloud feature of these sub-trajectory, the feature representation ability of the overall point cloud map can be improved, thereby improving the accuracy and robustness of the point cloud map; in addition, by only focusing on the sub-trajectory that meets the curvature condition, the processing of low curvature sub-trajectory (such as straight driving trajectory) can be reduced, the computational complexity can be reduced, and the real-time performance of the system can be improved.

[0097] In one embodiment, the target type includes a weak texture type, and the terminal performs type detection on the sub-trajectory of the motion trajectory, and the process of obtaining the target sub-trajectory belonging to the target type specifically includes the following steps: performing point cloud distribution processing on a subset of the point cloud data of the sub-trajectory to be processed in the point cloud data set to obtain the normally distributed point cloud features of the sub-trajectory to be processed; based on the normally distributed point cloud features of the sub-trajectory to be processed, aligning the point cloud data of the sub-trajectory to be processed with the point cloud map to obtain the alignment success rate; in the motion trajectory, determining the sub-trajectory corresponding to the alignment success rate that meets the alignment conditions as the target sub-trajectory of the weak texture type.

[0098] It should be noted that the type detection of the sub-trajectory is performed sub-trajectory by sub-trajectory. For example, the type detection is performed on the first sub-trajectory first, and then the type detection is performed on the second sub-trajectory, and so on until the type detection of all sub-trajectories of the motion trajectory is completed. For any sub-trajectory, when performing type detection on it, it is necessary to process each frame of point cloud data in the point cloud data subset corresponding to the sub-trajectory frame by frame. For example, the first frame of point cloud data in the point cloud data subset is processed first, and then the second frame of point cloud data in the point cloud data subset is processed, ..., the i-th frame of point cloud data is processed, ..., until the w frames of point cloud data in the point cloud data subset are processed.

[0099] Point cloud distribution processing refers to the distribution analysis and processing of point cloud data to obtain feature information of point cloud data and realize applications such as registration, classification, and segmentation of point cloud data. In the embodiment of the present application, point cloud distribution processing can specifically be NDT (Normal Distribution Transform) processing. The NDT algorithm is a commonly used point cloud registration and feature extraction method, which can be used for point cloud distribution processing to obtain feature information of point cloud data.

[0100] The sub-trajectory to be processed refers to the sub-trajectory that needs to be processed and analyzed. For example, the complete motion trajectory is divided into 10 sub-trajectories, of which the first two sub-trajectories have been type-detected, and the third sub-trajectory is the sub-trajectory to be processed.

[0101] The registration success rate refers to the ratio of the number of successfully matched features to the total number of point features during the registration process. This ratio can reflect the similarity between the point cloud data of the sub-track to be processed and the point cloud map. The higher the registration success rate, the better the matching degree between the sub-track to be processed and the point cloud map, and vice versa.

[0102] The registration condition may specifically pre-set a registration success rate threshold value, which is used to determine whether the sub-track is a target sub-track of a weak texture type. It can be understood that when the registration success rate of the sub-track is less than the registration success rate threshold value, the sub-track may be considered to meet the registration condition, and thus be considered as a target sub-track of a weak texture type.

[0103] Specifically, for any frame of point cloud data in the point cloud data subset of the sub-trajectory to be processed, the frame of point cloud data is the point cloud data to be processed in the point cloud data subset of the sub-trajectory to be processed, point cloud distribution processing is performed on the point cloud data to be processed, and the normal distribution point cloud features of the point cloud data to be processed are obtained. Based on the normal distribution point cloud features, the line features and surface features of the point cloud data to be processed are extracted, and the line features and surface features of the corresponding point cloud map are obtained. The line features and surface features of the point cloud data to be processed are matched with the line features and surface features of the corresponding point cloud map to obtain successful matching features. The number of successful features is calculated, and the ratio of the number of successful features to the total number of line features and surface features of the point cloud data to be processed is determined as the registration success rate of the point cloud data to be processed, until the registration success rate of each frame of point cloud data in the point cloud data subset is obtained, and the average value of the registration success rate of each frame of point cloud data is determined as the registration success rate of the sub-track to be processed, thereby obtaining the registration success rate of each sub-track to be processed, determining whether the registration success rate of each sub-track to be processed meets the registration conditions, and in each sub-track, determining the sub-track corresponding to the registration success rate that meets the registration conditions as the target sub-track of the weak texture type.

[0104] like Figure 5 The figure shows a frame of point cloud data in an embodiment, performing point cloud distribution processing on the frame of point cloud data to obtain the normal distribution point cloud features of the frame of point cloud data, extracting the line features and surface features of the frame of point cloud data based on the normal distribution point cloud features, and obtaining the line features and surface features of the corresponding point cloud map, matching the line features and surface features of the frame of point cloud data with the line features and surface features of the corresponding point cloud map, and obtaining Figure 6 The matching results shown in the figure show that the triangles filled with white represent surface features, and the triangles filled with black represent line features. Figure 6 The matching result shown in the figure splices the point cloud data of this frame into the corresponding point cloud map, and we can get Figure 7 The stitching effect shown.

[0105] In the above embodiment, the terminal performs point cloud distribution processing on a subset of the point cloud data of the sub-trajectory to be processed in the point cloud data set to obtain the normally distributed point cloud features of the sub-trajectory to be processed; based on the normally distributed point cloud features of the sub-trajectory to be processed, the point cloud data of the sub-trajectory to be processed is aligned with the point cloud map to obtain the alignment success rate; thereby, the target sub-trajectory of the weak texture type can be determined in the motion trajectory, and by distinguishing the weak texture sub-trajectories and extracting the first point cloud features and the second point cloud features from these sub-trajectories, the feature representation capability of the overall point cloud map can be improved, thereby improving the accuracy and robustness of the point cloud map; in addition, by focusing only on the weak texture sub-trajectories, the processing of the strong texture sub-trajectories can be reduced, the computational complexity can be reduced, and the real-time performance of the system can be improved.

[0106] In one embodiment, the target type includes a low-speed type, and the terminal performs type detection on the sub-trajectory of the motion trajectory, and the process of obtaining the target sub-trajectory belonging to the target type specifically includes the following steps: obtaining the vehicle driving speed corresponding to each sub-trajectory in the motion trajectory; in the motion trajectory, determining the sub-trajectory corresponding to the vehicle driving speed that meets the speed condition as the target sub-trajectory of the low-speed type.

[0107] The speed condition may specifically pre-set a vehicle speed threshold value, which is used to determine whether the sub-trajectory is a low-speed target sub-trajectory. It can be understood that when the vehicle speed of the sub-trajectory is less than the vehicle speed threshold value, the sub-trajectory can be considered to meet the speed condition, and thus it is regarded as a low-speed target sub-trajectory.

[0108] Specifically, the terminal determines the vehicle speed of each point on the sub-trajectory, and determines the vehicle speed of the corresponding sub-trajectory based on the vehicle speed of each point, and then determines whether the vehicle speed of the sub-trajectory meets the speed condition, and determines the sub-trajectory corresponding to the vehicle speed that meets the speed condition as a low-speed type target sub-trajectory.

[0109] Among them, for any sub-trajectory, the vehicle driving speed of the sub-trajectory is determined based on the vehicle driving speed of each point. Specifically, the average value of the vehicle driving speeds of each point is determined, and the obtained average value is determined as the vehicle driving speed of the sub-trajectory, or the maximum value of the vehicle driving speeds of each point is obtained and determined as the vehicle driving speed of the sub-trajectory.

[0110] In the above embodiment, the terminal obtains the vehicle driving speed corresponding to each sub-trajectory in the motion trajectory, so as to determine the target sub-trajectory of the low-speed type in the motion trajectory. By distinguishing the low-speed sub-trajectories and extracting the first point cloud features and the second point cloud features for these sub-trajectories, the feature representation capability of the overall point cloud map can be improved, thereby improving the accuracy and robustness of the point cloud map. In addition, by focusing only on the low-speed sub-trajectories, the processing of normal sub-trajectories (sub-trajectories traveling at normal speeds) can be reduced, the calculation complexity can be reduced, and the real-time performance of the system can be improved.

[0111] In one embodiment, the target type includes an initialization type, and the terminal performs type detection on a sub-track of the motion track to obtain a target sub-track belonging to the target type. Specifically, the process includes the following steps: obtaining a frame sequence corresponding to each sub-track in the motion track; and determining, in the motion track, a sub-track corresponding to a frame sequence that meets a frame sequence condition as a target sub-track of the initialization type.

[0112] The frame order condition may specifically pre-set a frame order threshold value, which is used to determine whether the sub-track is a target sub-track of the initialization type. It can be understood that when the frame order of the sub-track is less than or equal to the frame order threshold value, the sub-track is considered to meet the frame order condition, and is thus considered to be a target sub-track of the initialization type.

[0113] Specifically, the terminal obtains the frame sequence of the point cloud data frame corresponding to each point in the sub-track, and determines the maximum value in the frame sequence of each point as the frame sequence of the corresponding sub-track, and determines whether the frame sequence of the sub-track satisfies the frame sequence condition, and determines the sub-track corresponding to the frame sequence that satisfies the frame sequence condition as the target sub-track of the initialization type.

[0114] For example, if the multi-frame point cloud data corresponding to a sub-track is from the 1st frame to the 30th frame, then the frame sequence of the sub-track is confirmed to be 30, and the frame sequence threshold corresponding to the frame sequence condition is 30, then it is determined that the frame sequence of the sub-track satisfies the frame sequence condition, and the sub-track is determined to be the target sub-track of the initialization type.

[0115] In the above embodiment, the terminal obtains the frame sequence corresponding to each sub-track in the motion trajectory; determines the target sub-track of the initialization type in the motion trajectory, distinguishes the initialization sub-tracks, and extracts the first point cloud features and the second point cloud features of these sub-tracks, so as to improve the feature representation capability of the overall point cloud map, thereby improving the accuracy and robustness of the point cloud map; in addition, by focusing only on the initialization sub-tracks, the processing of ordinary sub-tracks (sub-tracks after initialization) can be reduced, the calculation complexity can be reduced, and the real-time performance of the system can be improved.

[0116] In one embodiment, the terminal aligns the point cloud data to be processed with the point cloud map based on the first point cloud feature and the second point cloud feature to obtain an updated point cloud map. The process specifically includes the following steps: determining the radar posture of the vehicle based on the first point cloud feature and the second point cloud feature; updating the point cloud data to be processed to the point cloud map based on the radar posture to obtain an updated point cloud map.

[0117] Among them, the radar posture of the vehicle refers to the posture of the laser radar configured on the vehicle. The posture refers to the position and attitude, which can be the position and attitude in the vehicle coordinate system or the position and attitude in the map coordinate system. It can be understood that the position and attitude in the vehicle coordinate system and the position and attitude in the map coordinate system can be converted to each other based on the transformation matrix.

[0118] Specifically, after obtaining the first point cloud feature and the second point cloud feature, the terminal can determine the first radar pose of the vehicle based on the first point cloud feature, and determine the second radar pose of the vehicle based on the second point cloud feature, and fuse the first radar pose with the second radar pose to obtain the radar pose of the vehicle.

[0119] In the above embodiment, the terminal can more accurately determine the radar pose of the vehicle by combining the first point cloud feature and the second point cloud feature. Based on the precise radar pose, the point cloud data to be processed can be more accurately updated to the point cloud map, thereby improving the quality of the point cloud map, making the point cloud map more detailed, complete and accurate, and providing a more reliable foundation for subsequent positioning, navigation, path planning and other tasks.

[0120] In one embodiment, the process of the terminal determining the radar posture of the vehicle based on the first point cloud feature and the second point cloud feature specifically includes the following steps: matching the first point cloud feature with the first point cloud data in the point cloud map to obtain the first radar posture; matching the second point cloud feature with the second point cloud data in the point cloud map to obtain the second radar posture; determining the radar posture of the vehicle based on the first radar posture and the second radar posture.

[0121] The first point cloud data is the reference point cloud data of the first registration algorithm, the first matching algorithm may be Frame-to-map matching, the first point cloud data may specifically be at least a part of the spliced ​​multi-frame point cloud data corresponding to the point cloud map, and the corresponding first radar pose is the pose of the vehicle's laser radar relative to the point cloud map; the second point cloud data is the reference frame point cloud data of the second registration algorithm, the second matching algorithm may be at least one of Frame-to-Frame matching and Frame-to-MultiFrames matching, and the second point cloud data may be At least a part of the processed multi-frame point cloud data corresponding to the point cloud map, the corresponding second radar posture is the posture of the vehicle's lidar relative to the reference frame point cloud data; the first point cloud data and the second point cloud data are partially the same, for example, the processed multi-frame point cloud data corresponding to the point cloud map is the 1st to 20th frame point cloud data in the point cloud data set, then the first point cloud data is the point cloud data obtained after splicing the 1st to 20th frames of point cloud data, and the second point cloud data is the first 5 frames of the point cloud data to be processed, then the second point cloud data is the 16th to 20th frame point cloud data in the point cloud data set.

[0122] In addition, the first point cloud data can specifically be at least a part of the spliced ​​multi-frame point cloud data corresponding to the point cloud map, and the second point cloud data can also be at least a part of the processed multi-frame point cloud data corresponding to the current sub-track in the point cloud map. The first point cloud data and the second point cloud data can be the same or different.

[0123] In some embodiments, the point cloud data subset corresponding to the target sub-trajectory of the initialization type contains a total of Q frames of point cloud data, and the 1st to jth frames in the point cloud data subset have been processed and updated to the point cloud map. The first point cloud data is determined to be the point cloud data obtained by splicing the 1st to jth frames of the point cloud data in the point cloud data subset with the spliced ​​point cloud data again. The second point cloud data is the first f frames of the point cloud data to be processed, where f=min(G,j), G≤Q. For example, let Q=30, G=5, and the point cloud data subset corresponding to the current sub-trajectory contains a total of 30 frames of point cloud data. If the 1st to 3rd frames in the point cloud data subset have been processed and updated to the point cloud map, the first point cloud data is determined to be the point cloud data The first point cloud data is the point cloud data obtained by splicing the 1st to 3rd frames of the point cloud data in the point cloud data subset with the spliced ​​point cloud data again, and the second point cloud data is the first f=min(5,3)=3 frames of the point cloud data to be processed, that is, the second point cloud data is the point cloud data of the 1st to 3rd frames of the point cloud data subset. If the 1st to 20th frames in the point cloud data subset have been processed and updated to the point cloud map, it is determined that the first point cloud data is the point cloud data of the 1st to 20th frames in the point cloud data subset, and the point cloud data obtained by splicing the spliced ​​point cloud data again, the second point cloud data is the first f=min(5,20)=5 frames of the point cloud data to be processed, that is, the second point cloud data is the point cloud data of the 16th to 20th frames of the point cloud data subset.

[0124] In some embodiments, the point cloud data subset corresponding to other sub-trajectories other than the initialization type in the target type contains a total of Q frames of point cloud data, and the 1st frame to the jth frame in the point cloud data subset has been processed and updated to the point cloud map. The first point cloud data is determined to be the 1st frame to the jth frame of point cloud data in the point cloud data subset, and the second point cloud data is the first f frames of the point cloud data to be processed, f=G. For example, let Q=30, G=5, and the point cloud data subset corresponding to the current target sub-trajectory (non-initialization type) contains a total of 30 frames of point cloud data. If the 1st frame to the 3rd frame in the point cloud data subset has been processed and updated to the point cloud map, the first point cloud data is determined to be the point of the processed sub-trajectory. Each frame of point cloud data in the cloud data set, and the 1st to 3rd frames of point cloud data in the point cloud data subset of the current sub-track, the second point cloud data is the first f=5 frames of the point cloud data to be processed, that is, the second point cloud data is the last two frames of point cloud data of the point cloud data subset of the previous sub-track, and the 1st to 3rd frames of point cloud data of the point cloud data subset of the current sub-track. If the 1st to 20th frames in the point cloud data subset have been processed and updated to the point cloud map, it is determined that the first point cloud data is the 1st to 20th frames of point cloud data in the point cloud data subset, and the second point cloud data is the first f=5 frames of the point cloud data to be processed, that is, the second point cloud data is the 16th to 20th frames of point cloud data in the point cloud data subset.

[0125] Specifically, the terminal determines the first point cloud data corresponding to the point cloud map, obtains the first historical point cloud feature obtained by performing point cloud distribution processing on the first point cloud data, matches the first point cloud feature with the first historical point cloud feature to obtain a first matching result, and determines the first radar pose based on the first matching result; the terminal determines the second point cloud data corresponding to the point cloud map, obtains the second historical point cloud feature obtained by performing geometric feature extraction on the second point cloud data, matches the second point cloud feature with the second historical point cloud feature to obtain a second matching result, and determines the second radar pose based on the second matching result; after obtaining the first radar pose and the second radar pose, the first radar pose and the second radar pose are fused according to a preset fusion algorithm to obtain the radar pose of the vehicle.

[0126] It can be understood that after each extraction of the first point cloud feature of a frame of point cloud data, the first point cloud feature can be stored as a historical point cloud feature in the first feature cache pool, so that when processing subsequent frames of point cloud data, the first historical point cloud feature corresponding to the first point cloud data can be directly selected from the cache pool for matching, avoiding repeated point cloud distribution processing on the same frame of point cloud data, reducing the amount of calculation and improving calculation efficiency. The first feature cache pool is used to cache the historical point cloud features of a specified number of frames. The size of the specified number of frames can be set according to actual conditions. For example, it can be set to be the same as the number of point cloud data frames corresponding to the point cloud map. When the number of frames of historical point cloud features in the first feature cache pool reaches the specified number of frames, the first feature cache pool is maintained in a first-in-first-out manner, that is, each time the historical point cloud features in the first feature cache pool are updated, the earliest frame of historical point cloud features is deleted before adding new historical point cloud features.

[0127] After each extraction of the second point cloud feature of a frame of point cloud data, the second point cloud feature can be stored as a historical point cloud feature in the second point cloud feature queue, so that when processing subsequent frames of point cloud data, the second point cloud feature of the subsequent frame point cloud data can be directly matched with the historical point cloud features in the queue frame by frame, avoiding repeated geometric feature extraction of the same frame of point cloud data, reducing the amount of calculation and improving calculation efficiency. The second feature queue is used to cache the historical point cloud features of a specified number of frames. The size of the specified number of frames can be set according to actual conditions, such as being set to the same as the reference frame number f. When the number of frames of the historical point cloud features in the second point cloud feature queue reaches the specified number of frames, each time the historical point cloud features in the second point cloud feature queue are updated, the historical point cloud features of the last frame of the second point cloud feature queue are first removed, and the historical point cloud features of the new frame are pushed into the head of the second point cloud feature queue.

[0128] In the above embodiment, the terminal obtains a first radar pose by matching the first point cloud data in the point cloud map based on the first point cloud feature, and obtains a second radar pose by matching the second point cloud data in the point cloud map based on the second point cloud feature. The radar pose of the vehicle is determined based on the first radar pose and the second radar pose. The two matching results can be comprehensively considered to reduce the possibility of mismatching, and the radar pose of the vehicle can be determined more accurately. Based on the precise radar pose, the point cloud data to be processed can be more accurately updated to the point cloud map, thereby improving the quality of the point cloud map, making the point cloud map more detailed, complete and accurate, and providing a more reliable foundation for subsequent positioning, navigation, path planning and other tasks.

[0129] In one embodiment, the terminal matches the first point cloud data in the point cloud map based on the first point cloud feature to obtain the first radar pose, which includes the following process: matching the first point cloud feature with the first point cloud data in the point cloud map to obtain the first pairing point; and determining the first radar pose based on the coordinates of the first pairing point.

[0130] Specifically, the terminal determines the first point cloud data corresponding to the point cloud map, obtains the first historical point cloud feature obtained by performing point cloud distribution processing on the first point cloud data, matches the first point cloud feature with the first historical point cloud feature to obtain a first pairing point, and determines the first radar pose based on the coordinates of the first pairing point.

[0131] The first pairing point includes a feature point in the first point cloud feature and a feature point in the corresponding first historical point cloud feature.

[0132] In one embodiment, after determining the first pairing point, the terminal may use the following formula to determine the first radar pose:

[0133]

[0134] In the above formula, T Mi is the first radar pose, M indicates that there are M pairs of first pairing points, m indicates the mth pair of first pairing points among the M pairs of first pairing points, represents the feature point in the first historical point cloud feature of the mth pair of first paired points, represents the feature point of the first point cloud feature in the mth pair of first paired points, argmin represents minimization solution. In the embodiment of the present application, the minimization solution problem is solved in an iterative manner with the help of nonlinear optimization. Specifically, the Lie algebra expression T = exp(ξ ^ ), then the single error term The derivatives of the pose can be obtained using the Lie algebra model:

[0135]

[0136] Through continuous iteration, we can find the pose T that minimizes the error. ri , which is the optimal relative pose between the current frame and the reference frame.

[0137] In the above embodiment, the terminal matches the first point cloud data in the point cloud map based on the first point cloud feature to obtain the first pairing point, and determines the first radar pose based on the coordinates of the first pairing point. Therefore, when the sub-track is a non-target sub-track, the point cloud data to be processed can be directly aligned with the point cloud map based on the first radar pose to obtain an updated point cloud map, thereby avoiding complex calculations on these non-target sub-tracks and improving the real-time performance of the system; when the sub-track is a target sub-track, the first radar pose is fused with the second radar pose to more accurately determine the radar pose of the vehicle, and based on the precise radar pose, the point cloud data to be processed can be more accurately updated to the point cloud map, thereby improving the quality of the point cloud map, making the point cloud map more detailed, complete and accurate, and providing a more reliable foundation for subsequent positioning, navigation, path planning and other tasks.

[0138] In one embodiment, the terminal matches the second point cloud data in the point cloud map based on the second point cloud feature to obtain the second radar pose, including the following steps: matching the second point cloud feature with the second point cloud data in the point cloud map to obtain second pairing points; determining the second radar pose based on the coordinates of each second pairing point.

[0139] Specifically, the terminal determines the second point cloud data corresponding to the point cloud map, obtains the second historical point cloud features obtained by performing geometric feature extraction on the second point cloud data, matches the second point cloud features with the second historical point cloud features to obtain a second pairing point, and determines the second radar pose based on the coordinates of the second pairing point.

[0140] The second matching points include feature points in the second point cloud features and feature points in the corresponding second historical point cloud features.

[0141] In one embodiment, after determining the second pairing point, the terminal may use the following formula to determine the second radar pose:

[0142]

[0143] In the above formula, T ri is the second radar pose, N indicates that there are N pairs of second paired points, n indicates the nth pair of second paired points among the N pairs of second paired points, Represents the feature point in the second historical point cloud feature in the nth pair of second paired points, The feature point representing the second point cloud feature in the nth pair of second paired points, in the embodiment of the present application, is based on the initial pose of the point cloud data to be processed relative to the reference frame point cloud data. Thereby determining the second radar pose.

[0144] In the above embodiment, the terminal matches the second point cloud data in the point cloud map based on the second point cloud feature to obtain the second pairing point; based on the coordinates of each second pairing point, the second radar pose is determined, and the second radar pose is integrated with the first radar pose to more accurately determine the radar pose of the vehicle. Based on the precise radar pose, the point cloud data to be processed can be more accurately updated to the point cloud map, thereby improving the quality of the point cloud map, making the point cloud map more detailed, complete and accurate, and providing a more reliable foundation for subsequent positioning, navigation, path planning and other tasks.

[0145] In one embodiment, the process of the terminal determining the radar posture of the vehicle based on the first radar posture and the second radar posture specifically includes the following steps: determining a first weight corresponding to the first radar posture and a second weight corresponding to the second radar posture; determining the radar posture of the vehicle based on the first weight, the first radar posture, the second weight and the second radar posture.

[0146] Among them, the first weight and the second weight are weight factors introduced to fuse the pose estimation results from two different features, and they respectively represent the contribution of the first radar pose and the second radar pose in the final pose estimation.

[0147] It should be noted that the first weight and the second weight can be determined based on the reliability, accuracy or matching degree of the respective features. For example, if the matching degree of the first point cloud feature (such as the normal distribution point cloud feature) is high, the first weight can be set to be larger; correspondingly, if the matching degree of the second point cloud feature (such as the geometric feature) is high, the second weight can be set to be larger.

[0148] Specifically, the terminal obtains a first weight corresponding to the first radar posture and a second weight corresponding to the second radar posture, and based on the obtained first weight and second weight, fuses the first radar posture and the second radar posture by weighted averaging or other optimization methods to obtain the radar posture of the vehicle. Such a fusion method can make full use of the information of the two features and improve the accuracy and robustness of the vehicle's radar posture.

[0149] In the above embodiment, the terminal determines the first weight corresponding to the first radar posture and the second weight corresponding to the second radar posture, so as to perform weighted fusion of the first radar posture and the second radar posture based on the weights, thereby more accurately determining the radar posture of the vehicle. Based on the precise radar posture, the point cloud data to be processed can be more accurately updated to the point cloud map, thereby improving the quality of the point cloud map, making the point cloud map more detailed, complete and accurate, and providing a more reliable foundation for subsequent positioning, navigation, path planning and other tasks.

[0150] In one embodiment, the sub-trajectory corresponds to at least two frames of point cloud data; the process in which the terminal determines the first weight corresponding to the first radar posture and the second weight corresponding to the second radar posture specifically includes the following steps: obtaining the number of point cloud frames of the point cloud data subset corresponding to the sub-trajectory, and determining the processing order of the point cloud data to be processed in the point cloud data subset corresponding to the sub-trajectory; determining the ratio of the processing order to the point cloud frame number as the first weight; determining the difference between the point cloud frame number and the processing order; and determining the ratio of the difference to the point cloud frame number as the second weight.

[0151] It should be noted that the motion trajectory corresponds to a point cloud data set, and each sub-trajectory of the motion trajectory corresponds to a point cloud data subset. In the embodiment of the present application, the processing process of the point cloud data set can be specifically performed on a point cloud data subset basis, for example, the first point cloud data subset corresponding to the first sub-trajectory is processed first, and then the second point cloud data subset corresponding to the second sub-trajectory is processed, and so on until all the point cloud data in the point cloud data set are processed; for the point cloud data subset corresponding to any sub-trajectory, the process of processing each frame of point cloud data in the point cloud data subset can be performed frame by frame, for example, the first frame of point cloud data in the subset is processed first, and then the second frame of point cloud data is processed,..., the i-th frame of point cloud data is processed until the w frames of point cloud data in the point cloud data subset are processed.

[0152] Specifically, when the terminal processes any frame of point cloud data, the frame of point cloud data is the point cloud data to be processed, then the sub-track corresponding to the point cloud data to be processed is determined, and the number of point cloud frames in the point cloud subset corresponding to the sub-track is obtained, as well as the processing order of the point cloud data to be processed in the point cloud subset corresponding to the sub-track, and the ratio of the processing order to the point cloud frame number is determined as the first weight; the difference between the point cloud frame number and the processing order is determined; and the ratio of the difference to the point cloud frame number is determined as the second weight.

[0153] For example, the point cloud data set corresponding to the motion trajectory contains 900 frames of point cloud data. The point cloud data to be processed currently is the 65th frame. The 65th frame of point cloud data corresponds to the third sub-trajectory of the motion trajectory. The point cloud data subset corresponding to the third sub-trajectory is [61 frames, 90 frames] point cloud data. The number of point cloud frames in this point cloud data subset is 30. The processing order of the 65th frame in this point cloud data subset is 5. Therefore, the first weight is determined to be 5 / 30, and the second weight is determined to be (30-5) / 30.

[0154] In one embodiment, the radar pose of the vehicle may be determined using the following formula:

[0155]

[0156] in, is the radar pose of the vehicle, is the first weight, is the second weight.

[0157] In the above embodiment, the terminal obtains the number of point cloud frames of the point cloud data subset corresponding to the sub-trajectory, and determines the processing order of the point cloud data to be processed in the point cloud data subset corresponding to the sub-trajectory; determines the ratio of the processing order to the point cloud frame number as the first weight; determines the difference between the point cloud frame number and the processing order; and determines the ratio of the difference to the point cloud frame number as the second weight, so that the first weight and the second weight can be dynamically adjusted according to the processing order of the point cloud data to be processed in the sub-trajectory, so as to adapt to the changes in the characteristics of the point cloud data at different times and locations, and the radar posture of the vehicle can be determined more accurately. Based on the precise radar posture, the point cloud data to be processed can be more accurately updated to the point cloud map, thereby improving the quality of the point cloud map, making the point cloud map more detailed, complete and accurate, and providing a more reliable foundation for subsequent positioning, navigation, path planning and other tasks.

[0158] In one embodiment, the terminal updates the point cloud data to be processed to the point cloud map based on the radar posture, and the process of obtaining the updated point cloud map specifically includes the following steps: obtaining the number of point cloud data frames corresponding to the point cloud map; when the number of frames reaches a frame number threshold, deleting the target frame point cloud data in the map point cloud data to obtain a deleted point cloud map; updating the point cloud data to be processed to the deleted point cloud map based on the radar posture to obtain an updated point cloud map.

[0159] It should be noted that with the processing of each frame of point cloud data in the point cloud dataset, or with the accumulation of processed sub-trajectories, for example, the point cloud data corresponding to the point cloud map exceeds 100 frames or the cumulative mileage of the sub-trajectory reaches 30m, the point cloud data in the point cloud dataset is continuously updated to the point cloud map, so that the amount of point cloud data corresponding to the point cloud map gradually increases, thereby affecting the efficiency of subsequent updates to the point cloud map; or the geographical range corresponding to the point cloud map is too large, such as the geographical range corresponding to the point cloud map is larger than a sphere with a radius of 100m, resulting in reduced accuracy of positioning and map matching.

[0160] Specifically, the terminal can use a queue data structure to store each frame of point cloud data corresponding to the point cloud map, and determine the number of frames of point cloud data stored in the queue. When the frame number reaches a preset frame number threshold, or the cumulative mileage corresponding to the point cloud map exceeds the mileage threshold, or the geographical range corresponding to the point cloud map is greater than the range threshold, the earliest frame of point cloud data in the queue is determined as the target frame point cloud data, and the target frame point cloud data is deleted from the queue to delete the target frame point cloud data from the point cloud map to obtain the deleted point cloud map, transform the point cloud data to be processed according to the radar posture, align the transformed point cloud data to be processed with the map coordinate system, and add the transformed point cloud data to be processed to the queue to add the transformed point cloud data to be processed to the deleted point cloud map to obtain an updated point cloud map.

[0161] In the above embodiment, the terminal obtains the number of point cloud data frames corresponding to the point cloud map, and when the number of frames reaches the frame number threshold, the target frame point cloud data in the map point cloud data is deleted to obtain the deleted point cloud map, and the point cloud data to be processed is updated to the deleted point cloud map based on the radar posture to obtain the updated point cloud map, thereby effectively controlling the data volume of the point cloud map, preventing the point cloud map from being too large and causing insufficient computing and storage resources, reducing the computing complexity, and speeding up the update speed of the point cloud map.

[0162] In one embodiment, after obtaining the updated point cloud map, the terminal can also obtain real-time point cloud data collected by the target vehicle; perform feature processing on the real-time point cloud data to obtain real-time point cloud features; and align the real-time point cloud data with the updated point cloud map based on the real-time point cloud features to obtain the position of the target vehicle in the updated point cloud map.

[0163] The target vehicle is a vehicle that is undergoing automatic navigation or automatic driving. The vehicle is equipped with a variety of sensor devices, such as laser radar, camera, GPS, IMU, etc., which can obtain the vehicle's position, posture, speed and other information in real time, and obtain point cloud data of the road environment in real time, that is, real-time point cloud data. Feature processing can specifically be point cloud distribution processing or geometric feature extraction, specifically NDT processing or LOAM extraction.

[0164] Specifically, real-time point cloud data of the surrounding environment is collected by a laser radar mounted on the target vehicle, and the terminal obtains the real-time point cloud data transmitted by the laser radar. The real-time collected point cloud data is preprocessed by removing noise, downsampling, etc. to obtain preprocessed real-time point cloud data, and then feature extraction methods such as NDT and LOAM are used to perform feature processing on the preprocessed real-time point cloud data to obtain real-time point cloud features. Then, a registration algorithm such as NDT and LOAM is used to match the real-time point cloud features with the map features of the updated point cloud map, so that the real-time point cloud data is registered with the updated point cloud map. The registration algorithm continuously optimizes the target vehicle's posture (position and direction) by minimizing the error between the real-time point cloud features and the map features until it converges to an acceptable error range, thereby outputting the position of the target vehicle in the updated point cloud map, which can be used to realize automatic navigation or automatic driving functions, such as route planning, obstacle avoidance, etc.

[0165] In the above embodiment, the terminal obtains real-time point cloud data collected by the target vehicle; performs feature processing on the real-time point cloud data to obtain real-time point cloud features; and aligns the real-time point cloud data with the updated point cloud map based on the real-time point cloud features, so that the position of the target vehicle in the updated point cloud map can be calculated in real time, thereby improving positioning accuracy and providing real-time and accurate position information for vehicle navigation, path planning and obstacle avoidance.

[0166] In one embodiment, Figure 8 As shown, a method for processing a point cloud map is also provided, and the method is applied to Figure 1 The computer device in the example is used to illustrate, including the following steps:

[0167] S802, obtaining a motion trajectory of the vehicle and a point cloud data set collected according to the motion trajectory.

[0168] S804, determining the curvature value of each sub-trajectory in the motion trajectory; in the motion trajectory, determining the sub-trajectory corresponding to the curvature value that meets the curvature condition as a target sub-trajectory of the turning type.

[0169] S806, performing point cloud distribution processing on a subset of the point cloud data of the sub-trajectory to be processed in the point cloud data set to obtain the normally distributed point cloud features of the sub-trajectory to be processed; based on the normally distributed point cloud features of the sub-trajectory to be processed, registering the point cloud data of the sub-trajectory to be processed with the point cloud map to obtain the registration success rate; in the motion trajectory, determining the sub-trajectory corresponding to the registration success rate that meets the registration conditions as the target sub-trajectory of the weak texture type.

[0170] S808, obtaining the vehicle driving speed corresponding to each sub-trajectory in the motion trajectory; in the motion trajectory, determining the sub-trajectory corresponding to the vehicle driving speed that meets the speed condition as a low-speed type target sub-trajectory.

[0171] S810, obtaining a frame sequence corresponding to each sub-trajectory in the motion trajectory; in the motion trajectory, determining a sub-trajectory corresponding to a frame sequence that meets a frame sequence condition as a target sub-trajectory of an initialization type.

[0172] S812, performing point cloud distribution processing on the point cloud data to be processed in the point cloud data set to obtain a first point cloud feature of the point cloud data to be processed.

[0173] S814: When the sub-track corresponding to the point cloud data to be processed is not the target sub-track, the point cloud data to be processed is registered with the point cloud map based on the first point cloud feature to obtain an updated point cloud map.

[0174] S816, when the sub-trajectory corresponding to the point cloud data to be processed is the target sub-trajectory, geometric features are extracted from the point cloud data to be processed to obtain a second point cloud feature of the point cloud data to be processed; based on the first point cloud feature and the second point cloud feature, the point cloud data to be processed is aligned with the point cloud map to obtain an updated point cloud map.

[0175] The present application also provides an application scenario, which applies the above-mentioned point cloud map processing method. The point cloud map processing method specifically includes the following steps:

[0176] 1. Type detection of sub-trajectories of motion trajectories

[0177] The collection vehicle collects a large amount of GPS / IMU information and laser point cloud data at various households, determines the motion trajectory during collection based on the GPS / IMU information, and organizes the collected laser point cloud data according to the motion trajectory to obtain the corona data set; divides the motion trajectory into sub-trajectories, and performs type detection on each sub-trajectory to obtain target sub-trajectories of the initialization type, target sub-trajectory of the low-speed type, target sub-trajectory of the turning type, and target sub-trajectory of the weak texture type.

[0178] 2. Fusion SLAM of LOAM and NDT

[0179] If the sub-track corresponding to the processed current frame point cloud data is not the target sub-track, the NDT features are directly extracted from the current frame point cloud data, and the current frame point cloud data is aligned with the NDT map using the Frame-to-Map matching algorithm based on the NDT features to obtain an updated NDT map.

[0180] If the sub-track corresponding to the point cloud data of the current frame is the target sub-track, refer to Fig. 9As shown in the fusion SLAM schematic diagram, on the one hand, NDT features are extracted from the current frame point cloud data, and the NDT radar pose of the vehicle is determined by the Frame-to-Map matching algorithm based on the NDT features. On the other hand, LOAM features are extracted from the current frame point cloud data, and the LOAM radar pose of the vehicle is determined by the Frame-to-Frame matching algorithm based on the LOAM features. The NDT weight and LOAM weight corresponding to the current frame point cloud data are determined, and the NDT radar pose and LOAM radar pose are integrated based on the NDT weight and LOAM weight to obtain the fusion radar pose of the vehicle. The current frame point cloud data is aligned with the NDT map based on the fusion radar pose of the vehicle to obtain an updated NDT map.

[0181] 3. Map application

[0182] After obtaining the updated NDT map, the updated NDT map can be used to locate the autonomous driving vehicle, generate the driving trajectory of the autonomous driving vehicle based on the positioning result, and provide real-time and accurate location information for autonomous driving vehicle navigation, path planning and obstacle avoidance.

[0183] 4. Test results

[0184] Fig.10 and Fig.11 This is a schematic diagram of the positioning trajectory test results in an embodiment, which shows the trajectories obtained after SLAM based on the conventional NDT algorithm (origin) and based on the fusion of LOAM and NDT of the present application (new). GT (ap60) represents the trajectory collected by high-precision equipment (centimeter-level accuracy) as the reference true value to measure the algorithm accuracy. It can be seen from the figure that whether it is a straight trajectory or a turning trajectory, the fusion SLAM algorithm of LOAM and NDT of the present application is obviously closer to the reference trajectory.

[0185] Fig.12 and Fig.13 This is a schematic diagram of the point cloud stitching test results in an embodiment, which shows the point cloud stitching effects obtained after SLAM based on the conventional NDT algorithm (origin) and based on the fusion of LOAM and NDT of the present application (new). It can be seen from the figure that there is a large error in the trajectory stitching point cloud after SLAM based on the conventional NDT algorithm, while the fusion SLAM of LOAM and NDT of the present application accurately calculates the position and posture of the laser radar during the driving process of the collection vehicle, so the radar point cloud data can be accurately stitched on this basis.

[0186] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0187] Based on the same inventive concept, the embodiment of the present application also provides a point cloud map processing device for implementing the point cloud map processing method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the one or more point cloud map processing device embodiments provided below can refer to the limitations of the point cloud map processing method above, and will not be repeated here.

[0188] In one embodiment, Fig.14 As shown, a point cloud map processing device is provided, including: a collection data acquisition module 1402, a type detection module 1404, a first point cloud feature extraction module 1406, a first point cloud feature extraction module 1408 and a map update module 1410, wherein:

[0189] The acquisition data acquisition module 1402 is used to acquire the motion trajectory of the vehicle and the point cloud data set acquired according to the motion trajectory;

[0190] A type detection module 1404 is used to perform type detection on the sub-trajectories of the motion trajectory to obtain target sub-trajectories belonging to the target type;

[0191] The first point cloud feature extraction module 1406 is used to perform point cloud distribution processing on the point cloud data to be processed in the point cloud data set to obtain the first point cloud feature of the point cloud data to be processed;

[0192] A second point cloud feature extraction module 1408 is used to extract geometric features of the point cloud data to be processed to obtain a second point cloud feature of the point cloud data to be processed when the sub-track corresponding to the point cloud data to be processed is a target sub-track;

[0193] The map updating module 1410 is used to align the point cloud data to be processed with the point cloud map based on the first point cloud feature and the second point cloud feature to obtain an updated point cloud map.

[0194] In the above embodiment, after obtaining the motion trajectory of the vehicle and the point cloud data set collected according to the motion trajectory, the sub-trajectory of the motion trajectory is detected by type to obtain a target sub-trajectory belonging to the target type (such as a sub-trajectory corresponding to a scene where the existing point cloud SLAM algorithm has insufficient processing capabilities). For the target sub-trajectory of the target type, on the one hand, point cloud distribution processing is performed on the point cloud data to be processed of the target sub-trajectory to obtain a first point cloud feature of the point cloud data to be processed; on the other hand, geometric feature extraction is performed on the point cloud data to be processed of the target sub-trajectory to obtain a second point cloud feature of the point cloud data to be processed, thereby better extracting the features of the point cloud data to be processed, and fusing the first point cloud feature and the second point cloud feature to align the point cloud data to be processed with the point cloud map, thereby improving the accuracy and stability of the alignment, and thereby improving the generation speed and accuracy of the point cloud map.

[0195] In one embodiment, the target type includes a turning type; the type detection module 1404 is used to: determine the curvature value of each sub-track in the motion trajectory; in the motion trajectory, determine the sub-track corresponding to the curvature value that meets the curvature condition as the target sub-track of the turning type.

[0196] In one embodiment, the target type includes a weak texture type; the type detection module 1404 is used to: perform point cloud distribution processing on a subset of the point cloud data of the sub-trajectory to be processed in the point cloud data set to obtain the normally distributed point cloud features of the sub-trajectory to be processed; based on the normally distributed point cloud features of the sub-trajectory to be processed, align the point cloud data of the sub-trajectory to be processed with the point cloud map to obtain the alignment success rate; in the motion trajectory, determine the sub-trajectory corresponding to the alignment success rate that meets the alignment conditions as the target sub-trajectory of the weak texture type.

[0197] In one embodiment, the target type includes a low-speed type; the type detection module 1404 is used to: obtain the vehicle driving speed corresponding to each sub-track in the motion trajectory; in the motion trajectory, determine the sub-track corresponding to the vehicle driving speed that meets the speed condition as the target sub-track of the low-speed type.

[0198] In one embodiment, the target type includes an initialization type; the type detection module 1404 is used to: obtain the frame sequence corresponding to each sub-track in the motion trajectory; in the motion trajectory, determine the sub-track corresponding to the frame sequence that meets the frame sequence condition as the target sub-track of the initialization type.

[0199] In one embodiment, the map update module 1410 is used to: determine the radar posture of the vehicle based on the first point cloud feature and the second point cloud feature; update the point cloud data to be processed into the point cloud map based on the radar posture to obtain an updated point cloud map.

[0200] In one embodiment, the map update module 1410 is also used to: obtain a first radar pose based on matching the first point cloud feature with the first point cloud data in the point cloud map; obtain a second radar pose based on matching the second point cloud feature with the second point cloud data in the point cloud map; and determine the radar pose of the vehicle based on the first radar pose and the second radar pose.

[0201] In one embodiment, the map update module 1410 is further used to: match the first point cloud feature with the first point cloud data in the point cloud map to obtain a first pairing point; and determine the first radar pose based on the coordinates of the first pairing point.

[0202] In one embodiment, the map update module 1410 is further used to: match the second point cloud data in the point cloud map based on the second point cloud features to obtain second pairing points; and determine the second radar pose based on the coordinates of each second pairing point.

[0203] In one embodiment, the map update module 1410 is further used to: determine a first weight corresponding to the first radar pose, and a second weight corresponding to the second radar pose; and determine the radar pose of the vehicle based on the first weight, the first radar pose, the second weight, and the second radar pose.

[0204] In one embodiment, the point cloud data to be processed includes at least two frames of point cloud data; the map update module 1410 is further used to: obtain the number of point cloud frames of the point cloud data subset corresponding to the sub-trajectory, and determine the processing order of the point cloud data to be processed in the point cloud data subset corresponding to the sub-trajectory; determine the ratio of the processing order to the point cloud frame number as the first weight; determine the difference between the point cloud frame number and the processing order; and determine the ratio of the difference to the point cloud frame number as the second weight.

[0205] In one embodiment, the map update module 1410 is also used to: obtain the number of point cloud data frames corresponding to the point cloud map; when the number of frames reaches a frame number threshold, delete the target frame point cloud data in the map point cloud data to obtain a deleted point cloud map; based on the radar posture, update the point cloud data to be processed to the deleted point cloud map to obtain an updated point cloud map.

[0206] In one embodiment, Fig.15 As shown, the device also includes a real-time positioning module 1412, which is used to: obtain real-time point cloud data collected by the target vehicle; perform feature processing on the real-time point cloud data to obtain real-time point cloud features; and align the real-time point cloud data with the updated point cloud map based on the real-time point cloud features to obtain the position of the target vehicle in the updated point cloud map.

[0207] Each module in the above-mentioned point cloud map processing device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0208] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig.16 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface, the display unit and the input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for processing a point cloud map is implemented. The display unit of the computer device is used to form a visually visible image, and can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device casing, or an external keyboard, touchpad or mouse, etc.

[0209] Those skilled in the art will understand that Fig.16 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0210] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0211] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0212] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0213] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0214] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0215] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0216] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for processing a point cloud map, characterized in that: The method comprises: Acquire a motion trajectory of the vehicle and a point cloud data set collected according to the motion trajectory; Performing type detection on the sub-trajectories of the motion trajectory to obtain target sub-trajectories belonging to the target type; Performing point cloud distribution processing on the point cloud data to be processed in the point cloud data set to obtain a first point cloud feature of the point cloud data to be processed; When the sub-track corresponding to the point cloud data to be processed is the target sub-track, geometric feature extraction is performed on the point cloud data to be processed to obtain a second point cloud feature of the point cloud data to be processed; The point cloud data to be processed is registered with a point cloud map based on the first point cloud feature and the second point cloud feature to obtain an updated point cloud map.

2. The method according to claim 1, characterized in that The target type includes a turning type; and the type detection of the sub-trajectory of the motion trajectory to obtain a target sub-trajectory belonging to the target type includes: Determining the curvature value of each sub-track in the motion track; In the motion trajectory, a sub-trajectory corresponding to the curvature value that meets the curvature condition is determined as a target sub-trajectory of the turning type.

3. The method according to claim 1, characterized in that: The target type includes a weak texture type; and the performing type detection on the sub-track of the motion track to obtain a target sub-track belonging to the target type includes: Performing point cloud distribution processing on a point cloud data subset of the sub-trajectory to be processed in the point cloud data set to obtain a normally distributed point cloud feature of the sub-trajectory to be processed; Based on the normally distributed point cloud features of the sub-trajectory to be processed, registering the point cloud data of the sub-trajectory to be processed with the point cloud map to obtain a registration success rate; In the motion trajectory, a sub-trajectory corresponding to the registration success rate that meets the registration condition is determined as a target sub-trajectory of the weak texture type.

4. The method according to claim 1, characterized in that: The target type includes a low-speed type; and the type detection of the sub-trajectory of the motion trajectory to obtain a target sub-trajectory belonging to the target type includes: Obtaining the vehicle speed corresponding to each sub-track in the motion track; In the motion trajectory, a sub-trajectory corresponding to the vehicle driving speed that meets the speed condition is determined as the low-speed type target sub-trajectory.

5. The method according to claim 1, characterized in that The target type includes an initialization type; and the type detection of the sub-track of the motion track to obtain a target sub-track belonging to the target type includes: Obtaining a frame sequence corresponding to each sub-track in the motion track; In the motion trajectory, a sub-trajectory corresponding to the frame sequence that meets the frame sequence condition is determined as a target sub-trajectory of the initialization type.

6. The method according to claim 1, characterized in that The registering the to-be-processed point cloud data with the point cloud map based on the first point cloud feature and the second point cloud feature to obtain an updated point cloud map includes: Determining a radar pose of the vehicle based on the first point cloud features and the second point cloud features; The point cloud data to be processed is updated into the point cloud map based on the radar posture to obtain an updated point cloud map.

7. The method according to claim 6, characterized in that The determining the radar pose of the vehicle based on the first point cloud feature and the second point cloud feature includes: Obtaining a first radar pose based on matching the first point cloud feature with first point cloud data in the point cloud map; Obtaining a second radar pose based on matching the second point cloud feature with second point cloud data in the point cloud map; A radar pose of the vehicle is determined based on the first radar pose and the second radar pose.

8. The method according to claim 7, characterized in that The step of matching the first point cloud feature with the first point cloud data in the point cloud map to obtain a first radar pose includes: Matching the first point cloud feature with the first point cloud data in the point cloud map to obtain a first matching point; A first radar pose is determined based on the coordinates of the first paired points.

9. The method according to claim 7, characterized in that: The obtaining a second radar pose by matching the second point cloud feature with the second point cloud data in the point cloud map includes: Matching the second point cloud feature with the second point cloud data in the point cloud map to obtain a second matching point; A second radar pose is determined based on the coordinates of each of the second paired points.

10. The method according to claim 7, characterized in that The determining the radar pose of the vehicle based on the first radar pose and the second radar pose includes: Determine a first weight corresponding to the first radar posture and a second weight corresponding to the second radar posture; A radar pose of the vehicle is determined based on the first weight, the first radar pose, the second weight, and the second radar pose.

11. The method according to claim 10, characterized in that The sub-track corresponds to at least two frames of point cloud data; the determining of a first weight corresponding to the first radar posture and a second weight corresponding to the second radar posture includes: Acquire the number of point cloud frames of the point cloud data subset corresponding to the sub-trajectory, and determine the processing order of the point cloud data to be processed in the point cloud data subset corresponding to the sub-trajectory; Determine a ratio of the processing order to the number of point cloud frames as a first weight; Determining the difference between the point cloud frame number and the processing order; A ratio of the difference to the number of point cloud frames is determined as a second weight.

12. The method according to claim 6, characterized in that The updating of the to-be-processed point cloud data into the point cloud map based on the radar posture to obtain an updated point cloud map includes: Obtain the number of point cloud data frames corresponding to the point cloud map; When the number of frames reaches a frame number threshold, the target frame point cloud data in the map point cloud data is deleted to obtain a deleted point cloud map; The point cloud data to be processed is updated to the deleted point cloud map based on the radar posture to obtain an updated point cloud map.

13. The method according to any one of claims 1 to 12, characterized in that: The method further comprises: Obtain real-time point cloud data collected by the target vehicle; Performing feature processing on the real-time point cloud data to obtain real-time point cloud features; The real-time point cloud data is aligned with the updated point cloud map based on the real-time point cloud features to obtain the position of the target vehicle in the updated point cloud map.

14. A point cloud map processing device, characterized in that: The device comprises: A data acquisition module, used to acquire the motion trajectory of the vehicle and the point cloud data set acquired according to the motion trajectory; A type detection module, used for performing type detection on the sub-trajectory of the motion trajectory to obtain a target sub-trajectory belonging to a target type; A first point cloud feature extraction module is used to perform point cloud distribution processing on the point cloud data to be processed in the point cloud data set to obtain a first point cloud feature of the point cloud data to be processed; A second point cloud feature extraction module is used to extract geometric features of the point cloud data to be processed when the sub-track corresponding to the point cloud data to be processed is the target sub-track, so as to obtain a second point cloud feature of the point cloud data to be processed; A map updating module is used to align the point cloud data to be processed with the point cloud map based on the first point cloud feature and the second point cloud feature to obtain an updated point cloud map.

15. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 13 are implemented.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 13 are implemented.

17. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 13 are implemented.