Road boundary detection method, device and equipment

Abstract

The invention is suitable for the technical field of intelligent driving, and provides a road boundary detection method, device and equipment, and the method comprises the steps: obtaining the surrounding environment data of a current driving road, and the road point cloud data of a reference road corresponding to the current driving road, the surrounding environment data comprising laser radar point cloud data and image detection data; rasterizing the laser radar point cloud data, and counting height information of each point cloud to obtain height feature information corresponding to a mapping grid; based on the height feature information, performing first feature recognition on the laser radar point cloud data to obtain obstacle point cloud data; according to a mapping relation between the obstacle point cloud data and the image detection data, performing second feature recognition on the obstacle point cloud data to obtain boundary candidate point cloud data; and on the basis of the road point cloud data of the reference road, filtering the boundary candidate point cloud data to obtain actual road boundary data so as to improve the accuracy of road boundary detection.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of intelligent driving, and particularly relates to a road boundary detection method, device and equipment. BACKGROUND

[0002] In the field of intelligent driving, road boundary detection is a key technology to ensure the safe and stable driving of unmanned vehicles, which can obtain road boundaries in real time and accurately, and provide accurate navigation and positioning support for unmanned vehicles, thereby improving the driving safety and stability of unmanned vehicles.

[0003] In the prior art, the road boundary detection method mostly uses a single laser radar sensor and planned path information to identify obstacles around the vehicle, which cannot eliminate the interference of known obstacles such as people and vehicles, resulting in inaccurate road boundary data and difficulty in distinguishing road boundaries from obstacles in complex road scenes such as multiple turns and multiple irregularities. SUMMARY

[0004] The embodiments of the present application provide a road boundary detection method, device and equipment, which can solve the problem that in the prior art, the road boundary detection process uses a single laser radar sensor and planned path information to identify obstacles around the vehicle, which cannot eliminate the interference of known obstacles such as people and vehicles, resulting in inaccurate road boundary data and difficulty in distinguishing road boundaries from obstacles in complex road scenes such as multiple turns and multiple irregularities.

[0005] In a first aspect, the embodiments of the present application provide a road boundary detection method, which comprises:

[0006] obtaining peripheral environment data of a current driving road and road point cloud data of a reference road corresponding to the current driving road; wherein the peripheral environment data comprises laser radar point cloud data and image detection data;

[0007] performing rasterization processing on the laser radar point cloud data to obtain a plurality of mapping grids corresponding to the laser radar point cloud data;

[0008] statistically obtaining height feature information corresponding to each mapping grid from height information of point clouds in all laser radar point cloud data corresponding to each mapping grid;

[0009] performing first feature recognition on the laser radar point cloud data based on the height feature information corresponding to each mapping grid to obtain obstacle point cloud data;

[0010] performing second feature recognition on the obstacle point cloud data according to a mapping relationship between the obstacle point cloud data and the image detection data to obtain boundary candidate point cloud data;

[0011] Filter the boundary candidate point cloud data based on the road point cloud data of the reference road to obtain actual road boundary data.

[0012] In a possible implementation manner of the first aspect, the rasterizing the lidar point cloud data to obtain a plurality of mapping grids corresponding to the lidar point cloud data comprises:

[0013] Perform polar coordinate conversion on the coordinates of each point cloud in the lidar point cloud data to obtain point cloud converted coordinates corresponding to each point cloud;

[0014] Obtain a grid serial number of the mapping grid corresponding to each point cloud according to the point cloud converted coordinates corresponding to each point cloud and a maximum detection distance of a preset lidar;

[0015] Fill the lidar point cloud data corresponding to each point cloud according to the grid serial number of the mapping grid corresponding to each point cloud to obtain the mapping grid corresponding to each point cloud in the lidar point cloud data.

[0016] In a possible implementation manner of the first aspect, the performing first feature recognition on the lidar point cloud data based on the height feature information corresponding to each mapping grid to obtain obstacle point cloud data comprises:

[0017] Perform first feature recognition on the lidar point cloud data corresponding to each point cloud according to the height feature information corresponding to each mapping grid and a preset judgment condition to obtain ground point cloud data in the lidar point cloud data; wherein the height feature information at least comprises one of a maximum height value, a minimum height value, an average height value and a height difference value of point clouds in the mapping grid;

[0018] Filter the ground point cloud data in the lidar point cloud data to obtain the obstacle point cloud data.

[0019] In a possible implementation manner of the first aspect, the performing second feature recognition on the obstacle point cloud data according to a mapping relationship between the obstacle point cloud data and the image detection data to obtain boundary candidate point cloud data comprises:

[0020] Project each point cloud in the obstacle point cloud data into the image detection data to obtain a mapping relationship between each point cloud in the obstacle point cloud data and the image detection data; wherein the image detection data is dynamic obstacle data obtained by recognizing a dynamic obstacle on the current driving road.

[0021] Based on the mapping relationship between the obstacle point cloud data and the image detection data, the image detection pixel value corresponding to each point cloud in the obstacle point cloud data is obtained;

[0022] Based on preset recognition conditions, a second feature recognition is performed on the image detection pixel value corresponding to each point cloud in the obstacle point cloud data to determine whether each point cloud is the dynamic obstacle data.

[0023] If the image detection pixel value corresponding to the point cloud meets the preset recognition condition, then the point cloud is determined to be the dynamic obstacle data, and the obstacle point cloud data corresponding to the point cloud is deleted to obtain the boundary candidate point cloud data.

[0024] If the image detection pixel value corresponding to the point cloud does not meet the preset recognition conditions, then it is determined that the point cloud is not the dynamic obstacle point cloud data, and the obstacle point cloud data corresponding to the point cloud is retained to obtain the boundary candidate point cloud data.

[0025] In one possible implementation of the first aspect, projecting each point cloud in the obstacle point cloud data onto the image detection data to obtain a mapping relationship between each point cloud in the obstacle point cloud data and the image detection data includes:

[0026] The coordinates of each point cloud in the obstacle point cloud data are transformed using the camera coordinate system to obtain the camera transformed coordinates corresponding to each point cloud;

[0027] The camera transformation coordinates corresponding to each point cloud are mapped to the image detection data to obtain the mapping relationship between each point cloud in the obstacle point cloud data and the image detection data.

[0028] In one possible implementation of the first aspect, filtering the candidate boundary point cloud data based on the road point cloud data of the reference road to obtain the actual road boundary data includes:

[0029] The road point cloud data corresponding to all discrete path points on the benchmark road is converted into a corresponding binary tree structure, and the boundary candidate point cloud data is converted into an octree square voxel structure including multiple voxel center points.

[0030] The nearest neighbor searcher matches multiple voxel center points in the octree square voxel structure to multiple nearest discrete path points in the binary tree structure; wherein each nearest discrete path point corresponds to multiple voxel center points.

[0031] determine a nearest left voxel center point and a nearest right voxel center point from the plurality of voxel center points corresponding to each of the nearest discrete path points;

[0032] extract the boundary candidate point cloud data from a nearest left boundary candidate region corresponding to the nearest left voxel center point and a nearest right boundary candidate region corresponding to the nearest right voxel center point according to an octree voxel search method, to obtain actual road boundary data of the current driving road.

[0033] In a possible implementation manner of the first aspect, the determining of the nearest left voxel center point and the nearest right voxel center point from the plurality of voxel center points corresponding to each of the nearest discrete path points comprises:

[0034] transform the coordinates of the voxel center points according to a coordinate system in which each of the nearest discrete path points is located, to obtain voxel center point transformed coordinates corresponding to the voxel center points;

[0035] determine, according to the voxel center point transformed coordinates, the voxel center point with a minimum value in a positive direction from the nearest discrete path point as the nearest right voxel center point; and

[0036] determine, as the nearest left voxel center point, the voxel center point with a minimum value in a negative direction from the nearest discrete path point.

[0037] In a possible implementation manner of the first aspect, the obtaining of the road point cloud data of a reference road corresponding to the current driving road comprises:

[0038] obtaining, through positioning signals, road point cloud data of a plurality of roads at a distance of a preset threshold from the current driving road;

[0039] determining, according to the road point cloud data of the plurality of roads, one road closest to the current driving road from the plurality of roads as the reference road.

[0040] In a second aspect, an embodiment of the present application provides a road boundary detection device, and the device comprises:

[0041] a data acquisition module configured to acquire surrounding environment data of a current driving road and road point cloud data of a reference road corresponding to the current driving road; wherein the surrounding environment data comprises laser radar point cloud data and image detection data;

[0042] a grid processing module configured to perform grid processing on the laser radar point cloud data, to obtain a plurality of mapping grids corresponding to the laser radar point cloud data;

[0043] a height feature extraction module configured to statistically analyze height information of point clouds in all the laser radar point cloud data corresponding to each of the mapping grids to obtain height feature information corresponding to each of the mapping grids;

[0044] a first feature recognition module configured to perform first feature recognition on the laser radar point cloud data based on the height feature information corresponding to each of the mapping grids to obtain obstacle point cloud data;

[0045] a second feature recognition module configured to perform second feature recognition on the obstacle point cloud data according to a mapping relationship between the obstacle point cloud data and the image detection data to obtain boundary candidate point cloud data;

[0046] a road boundary determination module configured to filter the boundary candidate point cloud data based on road point cloud data of the reference road to obtain actual road boundary data.

[0047] In a third aspect, an embodiment of the present application provides a road boundary detection device, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the road boundary detection method of any one of the above aspects when executing the computer program.

[0048] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executable on a processor to implement the road boundary detection method of any one of the above aspects.

[0049] In a fifth aspect, an embodiment of the present application provides a computer program product, which, when executed on a terminal device, causes the terminal device to perform the road boundary detection method of any one of the above aspects.

[0050] Compared with the prior art, the embodiment of the present application has the following beneficial effects:

[0051] The road boundary detection method provided by the embodiment of the present application comprises the following steps: acquiring surrounding environment data of a current driving road and road point cloud data of a reference road corresponding to the current driving road; wherein the surrounding environment data comprises laser radar point cloud data and image detection data; first, the laser radar point cloud data is rasterized to obtain a plurality of mapping grids corresponding to the laser radar point cloud data; the height information of all point clouds in each mapping grid is counted to obtain height feature information corresponding to each mapping grid; second, based on the height feature information corresponding to each mapping grid, the laser radar point cloud data is subjected to first feature recognition to obtain obstacle point cloud data; then, according to the mapping relationship between the obstacle point cloud data and the image detection data, the obstacle point cloud data is subjected to second feature recognition to obtain boundary candidate point cloud data; finally, based on the road point cloud data of the reference road, the boundary candidate point cloud data is filtered to obtain actual road boundary data. The present application extracts the obstacle point cloud data by rasterizing the laser radar point cloud data; then, the image detection data is mapped into the obstacle point cloud data to eliminate dynamic point clouds, thereby obtaining the boundary candidate point cloud data; finally, the road point cloud data of the reference road of the high-precision map is used to extract the real actual road boundary data from the boundary candidate point cloud data, which can solve the problem that the prior art cannot eliminate the interference of known obstacles such as people and vehicles in the process of road boundary detection, and the identified road boundary data is inaccurate and difficult to distinguish between the road boundary and the obstacle in a complex road scene with multiple turns and irregularities, thereby improving the accuracy of actual road boundary data detection. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0053] Figure 1 is a flowchart of a road boundary detection method provided by an embodiment of the present application;

[0054] Figure 2 is a schematic diagram of image detection data provided by an embodiment of the present application;

[0055] Figure 3 is a schematic diagram of boundary candidate point cloud data matching to a reference road provided by an embodiment of the present application;

[0056] Figure 4is a flowchart of a road boundary detection method according to another embodiment of the present application;

[0057] Figure 5 is a schematic diagram of determining a nearest left voxel center point and a nearest right voxel center point according to an embodiment of the present application;

[0058] Figure 6 is a structural schematic diagram of a road boundary detection device according to an embodiment of the present application;

[0059] Figure 7 is a structural schematic diagram of a road boundary detection device according to an embodiment of the present application. DETAILED DESCRIPTION

[0060] In the following description, for purposes of explanation and not limitation, specific details are set forth, such as a particular sequence of actions, techniques, etc., in order to provide a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application can be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known methods, devices, and circuits are omitted so as not to obscure the description of the present application with unnecessary detail.

[0061] It is to be understood that the terminology "includes", "has", "holds", "contains" and / or "comprising", when used in this specification and in the following claims, indicates the presence of the described features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0062] It is also to be understood that the terminology "and / or" when used in this specification and in the following claims, refers to at least one of the items, or any combination of one or more of the items, and includes any possible combination of the items.

[0063] As used in this specification and in the claims, the terms "if" and "when" can be interpreted to mean "upon" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [a described condition or event] is detected" can be interpreted to mean "upon determining" or "in response to determining" or "upon detecting [the described condition or event]" or "in response to detecting [the described condition or event]", depending on the context.

[0064] In addition, in the description of the application and in the following claims, the terms "first", "second", "third", etc. are used only for distinguishing the description, and cannot be understood as indicating or implying relative importance.

[0065] Reference within the specification of this application to "one embodiment" or "some embodiments" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. The appearances of the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in additional embodiments," and so on, in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily referring to one or more particular embodiments that are "other than all embodiments." The terms "including," "comprising," "having," and variations thereof are meant to encompass the items listed thereafter, but do not exclude other items from also being present. Unless otherwise indicated, the terms "including," "comprising," and "having" are meant to encompass the items listed thereafter, but do not exclude other items from also being present.

[0066] Reference is made to Figure 1 , Figure 1 is a flowchart of a road boundary detection method provided by an embodiment of the application. The method comprises:

[0067] S11, acquiring surrounding environment data of a current driving road and road point cloud data of a reference road corresponding to the current driving road; wherein the surrounding environment data comprises laser radar point cloud data and image detection data;

[0068] S12, performing rasterization processing on the laser radar point cloud data to obtain a plurality of mapping grids corresponding to the laser radar point cloud data;

[0069] S13, performing statistics on height information of point clouds in all laser radar point cloud data corresponding to each mapping grid to obtain height feature information corresponding to each mapping grid;

[0070] S14, performing first feature recognition on the laser radar point cloud data based on the height feature information corresponding to each mapping grid to obtain obstacle point cloud data;

[0071] S15, performing second feature recognition on the obstacle point cloud data according to a mapping relationship between the obstacle point cloud data and the image detection data to obtain boundary candidate point cloud data;

[0072] S16, performing filtering on the boundary candidate point cloud data based on the road point cloud data of the reference road to obtain actual road boundary data.

[0073] It should be noted that in the present embodiment, the execution subject can be a terminal device such as a server, and no specific limitation is made in this regard.

[0074] In step S11, the surrounding environment data is the information of the current driving road and its surrounding environment obtained during the driving process of the vehicle, mainly including laser radar point cloud data and image detection data, which are used to analyze the traffic conditions and obstacle positions of the current driving road. The laser radar point cloud data is the three-dimensional point cloud data of the object surface on the driving road obtained by the laser radar on the vehicle, which contains a large number of three-dimensional coordinate points, i.e. point cloud, which can accurately describe the shape and position of the object. The image detection data is obtained by detecting and recognizing the two-dimensional image data obtained by the camera through a two-dimensional image detection model, and the dynamic obstacle data containing the dynamic obstacles of the vehicle and pedestrians on the current driving road is obtained. As shown in Figure 2 Figure 2 is a schematic diagram of the image detection data provided by an embodiment of the present application. The image detection data can be represented as a mask image, in which the recognized pedestrians, vehicles, roads, etc. can be filled with different pixel values. The image detection data can assist the laser radar point cloud data in more accurate road recognition and obstacle detection.

[0075] The reference road is a known standard road corresponding to the current driving road obtained by searching the positioning signal, and the road point cloud data of the reference road can be used as a reference for comparison with the laser radar point cloud data of the current road, so as to identify the road boundary of the current road, etc.

[0076] In step S12, the gridding processing refers to dividing the continuous three-dimensional space into a plurality of regular grids, and each grid contains a certain spatial range. The laser radar point cloud data is gridded to obtain a plurality of mapping grids corresponding to the laser radar point cloud data, and each mapping grid contains a plurality of point clouds. The laser radar point cloud data is the original scanning data of the surrounding environment of the current driving road, and the gridding processing of the laser radar point cloud data facilitates the statistics and analysis of the data. Specifically, the gridding processing is a projection of the three-dimensional laser radar point cloud data, that is, it is assumed that the ground is relatively flat, that is, the Z-axis direction of the point cloud corresponding to the ground object has small fluctuations, and the point cloud in the laser radar point cloud data is projected to the X-axis Y-axis grid plane by gridding the scanning area. That is, each point cloud (x, y, z) only retains its x and y coordinate information in the projection process, and the highest point, the lowest point, etc. of the Z-axis direction of each point cloud in the grid are counted to determine whether the corresponding point cloud in the grid is a ground point cloud or an obstacle point cloud. It can be seen that each point cloud contained in the mapping grid has a clear corresponding relationship with the point cloud in the laser radar point cloud data, and the subsequent data processing and analysis can be performed by counting and analyzing the characteristics of the point clouds contained in the mapping grid.

[0077] ​In step S13, height information refers to the height value of each point in each point cloud data relative to the ground. By statistically analyzing the height information of all LiDAR point cloud data corresponding to each mapping grid, the height feature information corresponding to each mapping grid can be obtained. By statistically analyzing the height feature information, features such as obstacles and road boundaries can be accurately identified.

[0078] The height feature information includes at least one of the following: the maximum height value, minimum height value, average height value, and height difference of the point cloud within the mapped raster. The maximum height value refers to the maximum value along the Z-axis among multiple point cloud data points in the mapped raster, reflecting the highest point of the object or terrain within that raster. The minimum height value refers to the minimum value along the Z-axis among multiple point cloud data points in the mapped raster, reflecting the lowest point of the object or terrain within that raster. The average height value refers to the average value along the Z-axis among multiple point cloud data points in the mapped raster, providing the average elevation of the object or terrain within that raster, eliminating the influence of individual extreme values ​​and reflecting the overall trend. The height difference refers to the difference between the maximum and minimum height values ​​along the Z-axis among multiple point cloud data points in the mapped raster, reflecting the degree of vertical undulation of the object or terrain within that raster.

[0079] In step S14, by analyzing the height feature information of the point cloud within each mapped grid, the LiDAR point cloud data undergoes its first feature recognition, yielding obstacle point cloud data. This first feature recognition is the preliminary feature identification of the LiDAR point cloud data based on the height feature information corresponding to the mapped grid. The data obtained after the first feature recognition is used as the obstacle point cloud data.

[0080] In step S15, by performing a second feature recognition on the obstacle point cloud data based on the mapping relationship between the obstacle point cloud data and the image detection data, boundary candidate point cloud data can be obtained. The second feature recognition further identifies features in the obstacle point cloud data based on the mapping relationship between the obstacle point cloud data and the image detection data. This second feature recognition improves the accuracy and reliability of obstacle detection. The data obtained after the second feature recognition is used as the boundary candidate point cloud data.

[0081] In step S16, the candidate point cloud data of the boundary obtained after the second feature recognition is filtered using the road point cloud data of the reference road to obtain the actual road boundary data. The actual road boundary data is the point cloud data of the real road boundary of the current driving road obtained after multiple recognitions and filterings of the LiDAR point cloud data. This enables accurate perception and recognition of the surrounding environment of the current driving road, providing reliable data support for intelligent driving.

[0082] It can be understood that the road boundary detection method provided by the embodiments of the present application obtains the surrounding environment data of the current driving road and the road point cloud data of the reference road corresponding to the current driving road, wherein the surrounding environment data includes laser radar point cloud data and image detection data. First, the laser radar point cloud data is rasterized to obtain a plurality of mapping grids corresponding to the laser radar point cloud data. The height information of the point cloud in all laser radar point cloud data corresponding to each mapping grid is counted to obtain the height feature information corresponding to each mapping grid. Second, based on the height feature information corresponding to each mapping grid, the laser radar point cloud data is subjected to first feature recognition to obtain obstacle point cloud data. Then, according to the mapping relationship between the obstacle point cloud data and the image detection data, the obstacle point cloud data is subjected to second feature recognition to obtain boundary candidate point cloud data. Finally, based on the road point cloud data of the reference road, the boundary candidate point cloud data is filtered to obtain actual road boundary data. The present application extracts obstacle point cloud data by rasterizing laser radar point cloud data. Then, combined with image detection data, the image detection data is mapped to the obstacle point cloud data to eliminate dynamic point cloud and obtain boundary candidate point cloud data. Finally, the actual road boundary data is extracted from the boundary candidate point cloud data by using the road point cloud data of the reference road of the high-precision map, which can solve the problem that the prior art cannot eliminate the interference of known obstacles such as people and vehicles in the process of road boundary detection by using a single laser radar sensor and planned path information to identify obstacles around the vehicle, leading to the problem that the identified road boundary data is inaccurate and difficult to distinguish between road boundaries and obstacles in complex road scenes such as multiple turns and irregularities, thereby improving the accuracy of actual road boundary data detection.

[0083] In a possible implementation, the rasterization of the laser radar point cloud data to obtain a plurality of mapping grids corresponding to the laser radar point cloud data includes:

[0084] The coordinates of each point cloud in the laser radar point cloud data are subjected to polar coordinate conversion to obtain the point cloud conversion coordinates corresponding to each point cloud.

[0085] According to the point cloud conversion coordinates corresponding to each point cloud and the maximum detection distance of the preset laser radar, the grid serial number of the mapping grid corresponding to each point cloud is obtained.

[0086] According to the grid serial number of the mapping grid corresponding to each point cloud, the laser radar point cloud data corresponding to each point cloud is filled to obtain the mapping grid corresponding to each point cloud in the laser radar point cloud data.

[0087] It should be noted that the coordinates of each point cloud in the laser radar point cloud data are subjected to polar coordinate conversion to obtain the point cloud conversion coordinates corresponding to each point cloud, that is, assuming that the coordinates of the point cloud are Ci (x i ,y i ,z i ), and polar coordinate conversion is performed on the point cloud to obtain point cloud conversion coordinates P i (r i ,a i ,z i ) corresponding to the point cloud. Then, the point cloud conversion coordinates P i (r i ,a i ,z i ) corresponding to the point cloud are as shown in formula (1-1).

[0088]

[0089] In formula (1-1), x i represents the value of the point cloud in the X-axis direction, y i represents the value of the point cloud in the Y-axis direction, z i represents the height value of the point cloud in the Z-axis direction, r i represents the distance of the point cloud to the origin of coordinates, and a i represents the rotation angle of the point cloud with respect to the Y-axis.

[0090] Then, according to the point cloud conversion coordinates corresponding to each point cloud and the maximum detection distance of the preset laser radar, the grid serial number of the mapping grid corresponding to each point cloud is obtained. The maximum detection distance of the laser radar refers to the farthest distance at which the sensor can detect an object. The maximum detection distance of the preset laser radar is a preset value, which is used to filter the point cloud in the mapping grid. Specifically, assuming that the maximum detection distance of the preset laser radar is r max , when r i >r max , the point cloud is removed, and the remaining point cloud can be calculated according to formula (1-2) to obtain the grid serial number corresponding to the point cloud.

[0091]

[0092] In formula (1-2), j represents the grid serial number corresponding to the mapping grid, △r represents the distance resolution of the mapping grid, △a represents the angle resolution of the mapping grid, r max represents the maximum detection distance of the preset laser radar.

[0093] Finally, according to the grid serial number of the mapping grid corresponding to each point cloud, the laser radar point cloud data corresponding to all point clouds is filled to obtain the mapping grid corresponding to each point cloud in the laser radar point cloud data.

[0094] In one possible implementation, based on the height feature information corresponding to each mapped grid, the lidar point cloud data undergoes a first feature recognition to obtain obstacle point cloud data, including:

[0095] Based on the height feature information corresponding to each mapped grid and the preset judgment conditions, the first feature recognition is performed on the lidar point cloud data corresponding to each point cloud to obtain the ground point cloud data in the lidar point cloud data; wherein, the height feature information includes at least one of the following: the maximum height value, minimum height value, average height value and height difference of the point cloud within the mapped grid.

[0096] By filtering the ground point cloud data in the lidar point cloud data, obstacle point cloud data is obtained.

[0097] It should be noted that after obtaining the mapping raster corresponding to each point cloud, the height information of all point clouds in each mapping raster is first statistically analyzed to obtain the height feature information corresponding to each mapping raster. The height feature information includes at least one of the following: the maximum height value, minimum height value, average height value, and height difference of the point clouds within the mapping raster. The maximum height value refers to the maximum value along the Z-axis among multiple point cloud data in the mapping raster, reflecting the highest point of the object or terrain within that mapping raster. The minimum height value refers to the minimum value along the Z-axis among multiple point cloud data in the mapping raster, reflecting the lowest point of the object or terrain within that mapping raster. The average height value refers to the average value along the Z-axis among multiple point cloud data in the mapping raster, providing the average elevation of the object or terrain within that mapping raster, which can eliminate the influence of individual extreme values ​​and reflect the overall trend. The height difference refers to the difference between the maximum and minimum height values ​​along the Z-axis among multiple point cloud data in the mapping raster, reflecting the degree of vertical undulation of the object or terrain within that mapping raster.

[0098] In this embodiment, for ease of explanation, parameter z is used. j Indicates in the mapping raster G j The height z of each point cloud within the cloud i The set of each mapped lattice G j Corresponding height feature information F j (d max ,d min ,d mean ,d diff As shown in formula (1-3).

[0099]

[0100] In formula (1-3), F j d represents the height feature information corresponding to the mapped raster. max d represents the maximum height value of the point cloud within the mapped raster.min represents the minimum height value of the point cloud in the mapping grid, d mean represents the average height value of the point cloud in the mapping grid, d diff represents the height difference value of the point cloud in the mapping grid.

[0101] Then, based on the height feature information corresponding to each mapping grid, the first feature recognition is performed on the laser radar point cloud data to obtain obstacle point cloud data, specifically:

[0102] According to the height feature information corresponding to each mapping grid and the preset judgment condition, the first feature recognition is performed on the laser radar point cloud data corresponding to each point cloud to obtain the ground point cloud data in the laser radar point cloud data. The preset judgment condition is a condition set when the first feature recognition is performed, which can identify the ground point cloud. When the point cloud C i (x i ,y i ,z i ) is subjected to the first feature recognition, the height value z i of the point cloud C i (x i ,y i ,z i ) is extracted, and the height value z i of the point cloud is classified and judged according to the preset judgment condition. The preset judgment condition is performed in the following order:

[0103] The first judgment condition is that the ground height threshold is set as △H, and when the height difference value d j of the mapping grid G diff ≤△H, all point clouds mapped to the mapping grid G j are regarded as ground point clouds;

[0104] The second judgment condition is that when the height value of the point cloud satisfies z i -d min ≤△H, the difference between the minimum height value of the current mapping grid G j and the minimum height value of the adjacent grid G k is further judged, that is, if G j d min -G k d min ≤△H, the point cloud is regarded as a ground point cloud; if G j d min -G k d min >△H, it needs to be continuously judged, and if z i -G k d min ≤2△H, the point cloud is regarded as a ground point cloud;

[0105] The third judgment condition is that when the height value of the point cloud satisfies z i -G j d min ≤2△H, it is further judged that the current mapping grid G j corresponding to the adjacent grid G k satisfies G k d diff ≤△H, if the proportion of the number of point clouds satisfying the requirement in the adjacent grid G k is greater than 0.5, the point cloud is regarded as a ground point cloud;

[0106] The fourth judgment condition is that when all the above conditions are not satisfied, the point cloud satisfying z i -G k d min ≤△H / 2 is regarded as a ground point cloud.

[0107] If none of the point clouds C i (x i ,y i ,z i ) satisfies the above preset judgment condition, the point cloud is regarded as obstacle point cloud data, that is, the ground point cloud data in the laser radar point cloud data is filtered, and the obstacle point cloud data can be obtained.

[0108] In a possible implementation, according to the mapping relationship between the obstacle point cloud data and the image detection data, the obstacle point cloud data is subjected to a second feature recognition to obtain boundary candidate point cloud data, including:

[0109] Each point cloud in the obstacle point cloud data is projected into the image detection data to obtain the mapping relationship between each point cloud in the obstacle point cloud data and the image detection data; wherein the image detection data is dynamic obstacle data obtained by recognizing a dynamic obstacle on a current driving road;

[0110] According to the mapping relationship between the obstacle point cloud data and the image detection data, the image detection pixel value corresponding to each point cloud in the obstacle point cloud data is obtained;

[0111] According to a preset recognition condition, the image detection pixel value corresponding to each point cloud in the obstacle point cloud data is subjected to a second feature recognition to determine whether each point cloud is dynamic obstacle data;

[0112] If the image detection pixel value corresponding to the point cloud satisfies the preset recognition condition, the point cloud is determined as dynamic obstacle data, and the obstacle point cloud data corresponding to the point cloud is deleted to obtain boundary candidate point cloud data;

[0113] If the image detection pixel value corresponding to the point cloud does not satisfy the preset identification condition, it is determined that the point cloud is not dynamic obstacle point cloud data, and the obstacle point cloud data corresponding to the point cloud is retained to obtain boundary candidate point cloud data.

[0114] It should be noted that projecting each point cloud in the obstacle point cloud data into the image detection data can obtain the mapping relationship between each point cloud in the obstacle point cloud data and the image detection data. Through the mapping relationship between the obstacle point cloud data and the image detection data, the image detection pixel value corresponding to each point cloud in the obstacle point cloud data can be obtained. Then, the second feature identification is performed on the image detection pixel value corresponding to each point cloud in the obstacle point cloud data according to the preset identification condition, to determine whether each point cloud is dynamic obstacle data. The preset identification condition is a condition that can identify boundary candidate point cloud data when performing the second feature identification on the obstacle point cloud data. Whether each point cloud is dynamic obstacle data is determined according to whether the image detection pixel value satisfies the preset identification condition, so as to obtain the boundary candidate point cloud data. Specifically, if the image detection pixel value corresponding to the point cloud satisfies the preset identification condition, it is determined that the point cloud is dynamic obstacle data, that is, the point cloud represents a dynamic obstacle such as a vehicle or a pedestrian, and the point cloud is deleted from the obstacle point cloud data, so as to obtain the boundary candidate point cloud data. If the image detection pixel value corresponding to the point cloud does not satisfy the preset identification condition, it is determined that the point cloud is not dynamic obstacle point cloud data, and the obstacle point cloud data corresponding to the point cloud is retained, so as to obtain the boundary candidate point cloud data.

[0115] In a possible implementation, projecting each point cloud in the obstacle point cloud data into the image detection data to obtain the mapping relationship between each point cloud in the obstacle point cloud data and the image detection data comprises:

[0116] Performing camera coordinate system conversion on the coordinates of each point cloud in the obstacle point cloud data to obtain the camera conversion coordinates corresponding to each point cloud.

[0117] Mapping the camera conversion coordinates corresponding to each point cloud into the image detection data to obtain the mapping relationship between each point cloud in the obstacle point cloud data and the image detection data.

[0118] It should be noted that in the process of projecting each point cloud in the obstacle point cloud data into the image detection data to obtain the mapping relationship between each point cloud in the obstacle point cloud data and the image detection data, the camera and the laser radar need to be calibrated by using the external parameters, so as to convert the point cloud C i (x i ,y i ,z i ) into the camera coordinate system to obtain the camera conversion coordinates corresponding to each point cloud The camera coordinate conversion formula is shown in formula (1-4).

[0119]

[0120] In formula (1-4), represents the point cloud C i (x i ,y i ,z i ) is converted into the corresponding camera conversion coordinates in the camera coordinate system represents a rotation matrix; represents a translation vector.

[0121] Further, the is mapped into the image detection data, and the mapping formula is shown in formula (1-5),

[0122]

[0123] In formula (1-5), λ represents a camera depth coefficient value, and if the depth relationship is not considered, λ can be regarded as 1; K(f x ,f y ,c x ,c y ) represents a camera intrinsic coefficient.

[0124] The camera conversion coordinates corresponding to each point cloud are mapped into the image detection data, and the mapping relationship between each point cloud in the obstacle point cloud data and the image detection data is obtained, that is, the image detection pixel value (u i ,v i ) corresponding to is obtained.

[0125] In a possible implementation, based on the road point cloud data of the reference road, the boundary candidate point cloud data is filtered to obtain actual road boundary data, including:

[0126] The road point cloud data corresponding to all discrete path points on the reference road is converted into a corresponding binary tree structure, and the boundary candidate point cloud data is converted into an octree square voxel structure including a plurality of voxel center points;

[0127] Through a proximity searcher, the plurality of voxel center points in the octree square voxel structure are matched to a plurality of nearest discrete path points in the binary tree structure; wherein each nearest discrete path point corresponds to a plurality of voxel center points;

[0128] From the plurality of voxel center points corresponding to each nearest discrete path point, a nearest left voxel center point and a nearest right voxel center point are determined;

[0129] Based on the octree voxel search method, boundary candidate point cloud data are extracted from the nearest left boundary candidate region corresponding to the nearest left voxel center point and the nearest right boundary candidate region corresponding to the nearest right voxel center point to obtain the actual road boundary data of the current driving road.

[0130] It should be noted that, firstly, a KDtree binary tree structure is established for the discrete points on the baseline road. To improve boundary extraction efficiency, the boundary candidate point cloud data is converted into an Octree octree square voxel structure. Specifically, the center point V of each voxel is... i It is considered as a square region. The KDtree binary tree structure is a data structure for partitioning K-dimensional data space. The Octree octree square voxel structure is a tree-like data structure used to describe three-dimensional space. Both KDtree and Octree are data structures for efficiently searching point cloud data; KDtree is more suitable for low-dimensional space and relatively dense point cloud data, while Octree is more suitable for handling sparse data and spatially uneven 3D point cloud data.

[0131] Secondly, by setting a preset road width threshold, voxel center points that do not meet the requirements are removed, and the remaining voxel center points V are... i The nearest discrete path point L on the corresponding baseline road is found using a nearest neighbor search. j This involves mapping the boundary candidate point cloud data of the entire octree square voxel structure to the nearest discrete path point set {L} of the baseline road. j Within}, where each nearest discrete path point L j It can contain multiple corresponding voxel centers V i ,like Figure 3 As shown. Figure 3 This is a schematic diagram illustrating the matching of boundary candidate point cloud data to a reference road according to an embodiment of this application. Figure 3 In the middle, O car Represents the vehicle coordinate system, V1-V 10 L1 represents the voxel center point, and L5 represents the nearest discrete path point. It can be seen that the nearest discrete path point L4 contains its corresponding voxel center points V4, V8, and V9.

[0132] Then, the nearest left voxel center point and the nearest right voxel center point are determined from a plurality of voxel center points corresponding to each nearest discrete path point; and the actual road boundary data of the current driving road is obtained by extracting boundary candidate point cloud data from a nearest left boundary candidate region corresponding to the nearest left voxel center point and a nearest right boundary candidate region corresponding to the nearest right voxel center point according to an octree voxel search method. The octree voxel search method is a tree data structure search method for describing a three-dimensional space, which recursively divides the three-dimensional space into voxels with the same attribute, and uses a tree structure to organize the voxels, thereby achieving efficient spatial search and data compression.

[0133] As shown in Figure 4 , Figure 4 is a flowchart of a road boundary detection method provided by another embodiment of the present application. It can be understood that the image detection data of the camera can effectively eliminate the interference of known obstacles such as dynamic people and vehicles, and extract more accurate road boundary data. The reference road point cloud data information of the high-precision map can also be used to eliminate the interference of vehicles and pedestrians under the blind area of the camera that cannot be recognized by the camera, and the left and right boundaries of the road can also be distinguished without the aid of the planned path line of the vehicle itself. The method can be applied to complex road scenes such as crossroads, large turns, and multi-lane roads.

[0134] In a possible implementation, the nearest left voxel center point and the nearest right voxel center point are determined from a plurality of voxel center points corresponding to each nearest discrete path point, comprising:

[0135] The coordinates of the voxel center points are converted according to the coordinate system in which each nearest discrete path point is located, to obtain voxel center point conversion coordinates corresponding to the voxel center points;

[0136] The voxel center point with the minimum distance to the positive direction of the nearest discrete path point is determined as the nearest right voxel center point according to the voxel center point conversion coordinates; and

[0137] The voxel center point with the minimum distance to the negative direction of the nearest discrete path point is determined as the nearest left voxel center point.

[0138] It should be noted that the voxel center points V j contained in each nearest discrete path point L i need to be further processed, i.e., the voxel center points V i (x i ,y i ) in the vehicle coordinate system are converted to the coordinate system corresponding to the nearest discrete path point L j to obtain the voxel center points V i (x i ,y i) corresponding to the voxel center point conversion coordinates wherein the voxel center point conversion formula is shown as formula (1-6):

[0139]

[0140] In formula (1-6), represents the voxel center point conversion coordinates θ represents the rotation angle between the nearest discrete path point L j and the vehicle coordinate system O car (x j , y j ) represents the translation vector relative to the vehicle coordinate system O car .

[0141] Then, according to the voxel center point conversion coordinates, the voxel center point with the minimum positive direction distance from the nearest discrete path point is determined as the nearest right voxel center point, and the voxel center point with the minimum negative direction distance from the nearest discrete path point is determined as the nearest left voxel center point. That is, the nearest left voxel center point and the nearest right voxel center point are determined according to the value of the voxel center point conversion coordinates .

[0142] As shown in Figure 5 , Figure 5 is a schematic diagram for determining the nearest left voxel center point and the nearest right voxel center point provided by an embodiment of the present application. As shown in Figure 5 , V1-V6 represent a plurality of voxel center points contained in the nearest discrete path point L j . First, define △t as a preset road width threshold, and according to the elimination condition , eliminate invalid voxel center points, that is, eliminate the voxel center point V4 in Figure 5 . Then, extract the voxel center point with the minimum positive and negative direction of the y-axis from the remaining voxel center points as the nearest boundary point, that is, take the voxel center point V3 in Figure 5 as the nearest left voxel center point of the nearest discrete path point L j , and take the voxel center point V5 in Figure 5 as the nearest right voxel center point of the nearest discrete path point L j .

[0143] In a possible implementation, the road point cloud data of the reference road corresponding to the current driving road is obtained, including:

[0144] The road point cloud data of a plurality of roads at a distance of a preset threshold from the current driving road is obtained through positioning signals;

[0145] According to the road point cloud data of the plurality of roads, one of the plurality of roads closest to the current driving road is determined as a reference road.

[0146] It should be noted that the road point cloud data of the plurality of roads within the preset threshold range from the current driving road is searched by using the positioning signal, and a path line closest to the current driving road of the vehicle is extracted as the reference road. In this embodiment, the data structure of the reference road can be represented as a set of discrete points with direction vectors with an interval of 0.5 m. The preset threshold is a value representing the distance, and the specific value of the preset threshold is not limited in this embodiment.

[0147] It should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0148] A road boundary detection method corresponding to the above embodiment, Figure 6 A structural schematic diagram of a road boundary detection device provided by an embodiment of the present application is shown, and only parts related to the embodiments of the present application are shown for ease of description.

[0149] Referring to Figure 6 The road boundary detection device 3 of this embodiment includes:

[0150] The data acquisition module 31 is configured to acquire the surrounding environment data of the current driving road and the road point cloud data of the reference road corresponding to the current driving road. The surrounding environment data includes laser radar point cloud data and image detection data.

[0151] The grid processing module 32 is configured to perform grid processing on the laser radar point cloud data to obtain a plurality of mapping grids corresponding to the laser radar point cloud data.

[0152] The height feature extraction module 33 is configured to statistically analyze the height information of the point cloud in all the laser radar point cloud data corresponding to each mapping grid to obtain height feature information corresponding to each mapping grid.

[0153] The first feature recognition module 34 is configured to perform first feature recognition on the laser radar point cloud data based on the height feature information corresponding to each mapping grid to obtain obstacle point cloud data.

[0154] The second feature recognition module 35 is configured to perform second feature recognition on the obstacle point cloud data according to the mapping relationship between the obstacle point cloud data and the image detection data to obtain boundary candidate point cloud data.

[0155] The road boundary determination module 36 is configured to filter the boundary candidate point cloud data based on the road point cloud data of the reference road to obtain actual road boundary data.

[0156] It should be noted that the information interaction between the modules in the road boundary detection device 3 described above, the execution process, and the like, are based on the same concept as the method embodiments of the present application. For specific functions and technical effects thereof, refer to the method embodiments part. Here, no further description is provided.

[0157] The embodiments of the present application further provide a road boundary detection device, as shown in Figure 7 Figure 7 FIG. 4 is a structural schematic diagram of a road boundary detection device according to an embodiment of the present application. Referring to FIG. 4, Figure 7 The road boundary detection device 4 of this embodiment includes a memory 41, a processor 42, and a computer program stored in the memory 41 and executable on the processor 42. When the processor 42 executes the computer program, the steps in any of the road boundary detection method embodiments described above are implemented.

[0158] The embodiments of the present application further provide a computer readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps in any of the method embodiments described above are implemented.

[0159] The embodiments of the present application further provide a computer program product. When the computer program product is executed on a mobile terminal, the mobile terminal is caused to implement the steps in any of the method embodiments described above.

[0160] ​The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. When the computer program is executed by a processor, the steps of each method embodiment described above can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms. The computer readable medium at least includes any entity or device capable of carrying the computer program code to the photographing device / terminal equipment, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium. For example, U disk, mobile hard disk, magnetic disk or optical disk, etc. In some jurisdictions, according to legislation and patent practice, the computer readable medium can not be an electrical carrier signal and a telecommunication signal.

[0161] In the above embodiments, the description of each embodiment has its own focus, and the parts not described or recorded in detail in a certain embodiment can be referred to the related description of other embodiments.

[0162] Those skilled in the art can appreciate that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0163] In the embodiments provided in the present application, it should be understood that the disclosed apparatus / network device and method can be implemented in other ways. For example, the apparatus / network device embodiments described above are only schematic. The division of the modules or units is only a logical function division, and there can be another division in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0164] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may also be distributed to multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiment scheme according to actual needs.

[0165] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A road boundary detection method characterized by, The method comprises the following steps: acquiring surrounding environment data of a current driving road and road point cloud data of a reference road corresponding to the current driving road; wherein the surrounding environment data comprises laser radar point cloud data and image detection data; performing rasterization processing on the laser radar point cloud data to obtain a plurality of mapping grids corresponding to the laser radar point cloud data; counting height information of point clouds in all the laser radar point cloud data corresponding to each mapping grid to obtain height feature information corresponding to each mapping grid; performing first feature recognition on the laser radar point cloud data based on the height feature information corresponding to each mapping grid to obtain obstacle point cloud data; performing second feature recognition on the obstacle point cloud data based on a mapping relationship between the obstacle point cloud data and the image detection data to obtain boundary candidate point cloud data; performing filtering on the boundary candidate point cloud data based on the road point cloud data of the reference road to obtain actual road boundary data.

2. The road boundary detection method according to claim 1, wherein The method of performing rasterization processing on the laser radar point cloud data to obtain a plurality of mapping grids corresponding to the laser radar point cloud data comprises the following steps: performing polar coordinate conversion on coordinates of each point cloud in the laser radar point cloud data to obtain point cloud conversion coordinates corresponding to each point cloud; obtaining a grid serial number of the mapping grid corresponding to each point cloud based on the point cloud conversion coordinates corresponding to each point cloud and a maximum detection distance of a preset laser radar; filling the laser radar point cloud data corresponding to each point cloud in the laser radar point cloud data according to the grid serial number of the mapping grid corresponding to each point cloud to obtain the mapping grid corresponding to each point cloud in the laser radar point cloud data.

3. The road boundary detection method according to claim 2, wherein The method of performing first feature recognition on the laser radar point cloud data based on the height feature information corresponding to each mapping grid to obtain obstacle point cloud data comprises the following steps: performing first feature recognition on the laser radar point cloud data corresponding to each point cloud based on the height feature information corresponding to each mapping grid and a preset judgment condition to obtain ground point cloud data in the laser radar point cloud data; wherein the height feature information at least comprises one of the following: a maximum height value, a minimum height value, an average height value and a height difference value of point clouds in the mapping grid; performing filtering on the ground point cloud data in the laser radar point cloud data to obtain the obstacle point cloud data.

4. The road boundary detection method according to claim 1, wherein The method of performing second feature recognition on the obstacle point cloud data based on a mapping relationship between the obstacle point cloud data and the image detection data to obtain boundary candidate point cloud data comprises the following steps: projecting each point cloud in the obstacle point cloud data into the image detection data to obtain a mapping relationship between each point cloud in the obstacle point cloud data and the image detection data; wherein the image detection data is dynamic obstacle data obtained by recognizing dynamic obstacles on the current driving road. According to the mapping relationship between the obstacle point cloud data and the image detection data, an image detection pixel value corresponding to each point cloud in the obstacle point cloud data is obtained; According to a preset identification condition, a second feature identification is performed on the image detection pixel value corresponding to each point cloud in the obstacle point cloud data, to determine whether each point cloud is the dynamic obstacle data; If the image detection pixel value corresponding to the point cloud satisfies the preset identification condition, it is determined that the point cloud is the dynamic obstacle data, and the obstacle point cloud data corresponding to the point cloud is deleted to obtain the boundary candidate point cloud data; If the image detection pixel value corresponding to the point cloud does not satisfy the preset identification condition, it is determined that the point cloud is not the dynamic obstacle point cloud data, and the obstacle point cloud data corresponding to the point cloud is retained to obtain the boundary candidate point cloud data.

5. The road boundary detection method according to claim 4, characterized by, The projection of each point cloud in the obstacle point cloud data into the image detection data to obtain the mapping relationship between each point cloud in the obstacle point cloud data and the image detection data includes: A camera coordinate system conversion is performed on the coordinates of each point cloud in the obstacle point cloud data to obtain a camera conversion coordinate corresponding to each point cloud; The camera conversion coordinate corresponding to each point cloud is mapped into the image detection data to obtain the mapping relationship between each point cloud in the obstacle point cloud data and the image detection data.

6. The road boundary detection method of claim 1, wherein, The filtering of the boundary candidate point cloud data based on the road point cloud data of the reference road to obtain actual road boundary data includes: The road point cloud data corresponding to all discrete path points on the reference road is converted into a corresponding binary tree structure, and the boundary candidate point cloud data is converted into an octree square voxel structure including a plurality of voxel center points; A plurality of voxel center points in the octree square voxel structure are matched to a plurality of nearest discrete path points in the binary tree structure by a proximity searcher; wherein each nearest discrete path point corresponds to a plurality of voxel center points; A nearest left voxel center point and a nearest right voxel center point are determined from a plurality of voxel center points corresponding to each nearest discrete path point; According to an octree voxel search method, the boundary candidate point cloud data is extracted from a nearest left boundary candidate region corresponding to the nearest left voxel center point and a nearest right boundary candidate region corresponding to the nearest right voxel center point to obtain actual road boundary data of the current driving road.

7. The road boundary detection method according to claim 6, wherein The determination of a nearest left voxel center point and a nearest right voxel center point from a plurality of voxel center points corresponding to each nearest discrete path point includes: According to a coordinate system in which each nearest discrete path point is located, the coordinates of the voxel center points are converted to obtain voxel center point conversion coordinates corresponding to the voxel center points; According to the voxel center point conversion coordinates, the voxel center point with the minimum distance in the positive direction from the nearest discrete path point is determined as the nearest right voxel center point; and, The voxel center point with the minimum negative direction distance to the nearest discrete path point is determined as a nearest left voxel center point.

8. The road boundary detection method of claim 1, wherein, The road point cloud data of the reference road corresponding to the current driving road is acquired, including: The road point cloud data of multiple roads at a distance of a preset threshold from the current driving road is acquired through positioning signals. One of the multiple roads that is closest to the current driving road is determined as the reference road according to the road point cloud data of the multiple roads.

9. A road boundary detection apparatus characterized by comprising: Including: A data acquisition module is configured to acquire surrounding environment data of a current driving road and road point cloud data of a reference road corresponding to the current driving road, wherein the surrounding environment data includes laser radar point cloud data and image detection data. A grid processing module is configured to perform grid processing on the laser radar point cloud data to obtain multiple mapping grids corresponding to the laser radar point cloud data. A height feature extraction module is configured to count height information of point clouds in all the laser radar point cloud data corresponding to each mapping grid to obtain height feature information corresponding to each mapping grid. A first feature identification module is configured to perform first feature identification on the laser radar point cloud data based on the height feature information corresponding to each mapping grid to obtain obstacle point cloud data. A second feature identification module is configured to perform second feature identification on the obstacle point cloud data according to a mapping relationship between the obstacle point cloud data and the image detection data to obtain boundary candidate point cloud data. A road boundary determination module is configured to filter the boundary candidate point cloud data based on the road point cloud data of the reference road to obtain actual road boundary data.

10. A road boundary detection apparatus characterized by comprising: The computer program stored in the memory and executable on the processor implements the method of any one of claims 1 to 8 when the processor executes the computer program.

Images