Ground plane detection method and device and computer storage medium

Abstract

The invention provides a ground plane detection method and device and a computer storage medium, and the method comprises the steps: projecting a ground plane point cloud to a preset fan-shaped grid, and generating a plurality of coarse grid point cloud clusters; obtaining the thickness and the central point distance of each coarse grid point cloud cluster; determining a rough grid point cloud cluster of the concave-convex target according to the thickness and the central point distance; dividing the rough grid point cloud cluster of the concave-convex target into a plurality of fine grid point cloud clusters; acquiring the curvature of adjacent points of each fine grid point cloud cluster; acquiring concave-convex target point cloud in the fine grid point cloud cluster according to the curvature of the adjacent points; and outputting the concave-convex target of the ground plane by using the concave-convex target point cloud. According to the method, the block domains of the ground plane are refined through point cloud rasterization, the ground plane is divided into the small block domains to find the ground defects, and the detection precision can be effectively guaranteed.

Description

Technical Field

[0001] The present application relates to the field of robotics, and in particular to a ground plane detection method, device and computer storage medium. Background Art

[0002] With the rapid development of the autonomous driving industry, more and more autonomous driving vehicles are coming into people's field of vision. The higher the level of autonomous driving capabilities, the more they rely on multimodal perception systems. Cameras and lidars, as the two mainstream sensors for autonomous driving, each have their own advantages and disadvantages. The fusion of the two has become a direction that is more recognized by the mainstream autonomous driving industry.

[0003] Existing algorithms for detecting potholes on the ground usually rely on image semantic segmentation models to identify potholes and bumps, and then use binocular cameras or lidar to sense the depth of the target. This strategy strictly relies on the ability of image semantic segmentation to identify potholes. It is usually difficult to identify small and medium-sized potholes with little difference in pixels, and it is also difficult to identify the undulations of the ground. Summary of the invention

[0004] The present application provides a ground plane detection method, device and computer storage medium.

[0005] To solve the above technical problems, the present application proposes a ground plane detection method, which includes: projecting a ground plane point cloud onto a preset fan-shaped grid to generate a number of coarse grid point cloud clusters; obtaining the thickness and center point distance of each of the coarse grid point cloud clusters; determining the concave-convex target coarse grid point cloud cluster according to the thickness and the center point distance; dividing the concave-convex target coarse grid point cloud cluster into a number of fine grid point cloud clusters; obtaining the adjacent point curvature of each of the fine grid point cloud clusters; obtaining the concave-convex target point cloud in the fine grid point cloud cluster according to the adjacent point curvature; and using the concave-convex target point cloud to output the concave-convex target of the ground plane.

[0006] Wherein, the obtaining of the thickness and center point distance of each of the coarse grid point cloud clusters comprises: determining the thickness of the coarse grid point cloud cluster based on the difference between the maximum ordinate value and the minimum ordinate value in each of the coarse grid point cloud clusters; obtaining the center point of the coarse grid point cloud cluster; determining the convex center point distance based on the difference between the maximum ordinate value and the ordinate value of the center point; determining the concave center point distance based on the difference between the minimum ordinate value and the ordinate value of the center point; determining the concave convex target coarse grid point cloud cluster according to the thickness and the center point distance comprises: determining the coarse grid point cloud cluster whose thickness is greater than a preset thickness threshold and the convex center point distance is greater than a preset convex threshold as the concave convex target coarse grid point cloud cluster; and determining the coarse grid point cloud cluster whose thickness is greater than a preset thickness threshold and the concave center point distance is greater than a preset concave threshold as the concave convex target coarse grid point cloud cluster.

[0007] Wherein, before obtaining the curvature of adjacent points of each of the fine grid point cloud clusters, the ground plane detection method further comprises: sorting the data points in the fine grid point cloud clusters according to horizontal distances.

[0008] Among them, obtaining the concave-convex target point cloud in the fine grid point cloud cluster according to the curvature of the adjacent points includes: when judging according to the curvature of the adjacent points that the fine grid point cloud cluster has a curvature increase or decrease change, obtaining the concave-convex target point cloud corresponding to the curvature change position.

[0009] Among them, the obtaining of the concave-convex target point cloud in the fine grid point cloud cluster according to the curvature of the adjacent points includes: when it is judged according to the curvature of the adjacent points that the boundary point cloud of the fine grid point cloud cluster has a monotonically increasing or monotonically decreasing curvature, the fine grid point cloud cluster is associated with the adjacent coarse grid point cloud cluster to obtain an associated point cloud cluster; when it is judged according to the curvature of the adjacent points that the associated point cloud cluster has an increasing or decreasing curvature, the concave-convex target point cloud corresponding to the curvature change position is obtained.

[0010] Among them, the projecting of the ground plane point cloud to a preset sector grid to generate a number of coarse grid point cloud clusters includes: obtaining the horizontal distance and horizontal heading angle of the data points of the ground plane point cloud; projecting the ground plane point cloud to a preset sector grid according to the horizontal distance and the horizontal heading angle of the data points to generate a number of coarse grid point cloud clusters.

[0011] Among them, the ground plane detection method also includes: obtaining a ground plane point cloud image, and obtaining a drivable area in the original image; mapping pixels of the drivable area to the ground plane point cloud image, and extracting the drivable area point cloud image; based on the drivable area point cloud image, extracting the ground plane point cloud from the original point cloud.

[0012] Wherein, obtaining the ground plane point cloud image includes: obtaining an original point cloud; converting the original point cloud to a camera coordinate system to obtain a ground plane camera point cloud; and converting the camera point cloud to an image coordinate system to obtain the ground plane point cloud image.

[0013] In order to solve the above technical problems, the present application proposes a ground plane detection device, which includes a memory and a processor coupled to the memory; wherein the memory is used to store program data, and the processor is used to execute the program data to implement the above ground plane detection method.

[0014] In order to solve the above technical problems, the present application proposes a computer storage medium, wherein the computer storage medium is used to store program data, and when the program data is executed by a computer, it is used to implement the above ground plane detection method.

[0015] Different from the prior art, the beneficial effects of the present application are as follows: the ground plane detection device projects the ground plane point cloud onto a preset fan-shaped grid to generate a number of coarse grid point cloud clusters; obtains the thickness and center point distance of each coarse grid point cloud cluster; determines the concave-convex target coarse grid point cloud cluster according to the thickness and the center point distance; divides the concave-convex target coarse grid point cloud cluster into a number of fine grid point cloud clusters; obtains the adjacent point curvature of each fine grid point cloud cluster; obtains the concave-convex target point cloud in the fine grid point cloud cluster according to the adjacent point curvature; and uses the concave-convex target point cloud to output the concave-convex target of the ground plane. The present application refines the block domain of the ground plane by point cloud rasterization, divides the ground plane into small blocks to find defects on the ground, and can effectively ensure the accuracy of detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 is a flow chart of a first embodiment of a ground plane detection method provided by the present application;

[0018] Figure 2 The ground plane detection method provided by this application Figure 1 Schematic diagram of the sub-steps of step S12;

[0019] Figure 3 is a structural schematic diagram of an embodiment of a ground plane detection device provided by the present application;

[0020] Figure 4It is a structural diagram of an embodiment of a computer storage medium provided by the present application. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] The ground plane detection method of the present application is applied to a ground plane detection device, wherein the ground plane detection device of the present application can be a server, or a system composed of a server and a local terminal. Accordingly, the various parts of the ground plane detection device, such as various units, subunits, modules, and submodules, can all be set in the server, or can be set in the server and the local terminal respectively.

[0023] Furthermore, the above-mentioned server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or it can be implemented as a single server. When the server is software, it can be implemented as multiple software or software modules, such as software or software modules used to provide distributed servers, or it can be implemented as a single software or software module, which is not specifically limited here. In some possible implementations, the ground plane detection method of the embodiment of the present application can be implemented by a processor calling a computer-readable instruction stored in a memory.

[0024] This application proposes a ground plane detection method, see Figure 1 , Figure 1 It is a flowchart of the first embodiment of the ground plane detection method provided by the present application.

[0025] like Figure 1 As shown, the specific steps are as follows:

[0026] Step S11: Projecting the ground plane point cloud onto a preset sector grid to generate a number of coarse grid point cloud clusters.

[0027] The coarse grid is a fan-shaped grid that is larger than the fine grid. The coarse grid is divided again to form the fine grid. The coarse grid point cloud cluster is a point cloud cluster formed by projecting the ground plane point cloud onto the coarse grid.

[0028] Specifically, in one embodiment of the present application, a method for generating a coarse grid point cloud cluster is proposed, specifically: a ground plane detection device obtains the horizontal distance and horizontal heading angle of the data points of the ground plane point cloud; the ground plane point cloud is projected to a preset fan-shaped grid according to the horizontal distance and horizontal heading angle of the data points to generate a plurality of coarse grid point cloud clusters.

[0029] Specifically, the present application uses a concentric circle model, and the ground plane detection device converts the ground plane point cloud P ground Project according to the BEV perspective (Bird's Eye View), divide the ground point cloud into 2D grids, and calculate the horizontal distance R between each point cloud data point xy ,in Where x and y are the coordinates of the data point.

[0030] The plane detection device further calculates the horizontal heading angle θ between each point cloud data point, θ = tan -1 (y / x), where x and y are the coordinates of the data point.

[0031] In the embodiment of the present application, a non-uniform sector-shaped grid is used to fully consider the sparse law of the point cloud and improve the detection accuracy.

[0032] When determining the size of the sector grid in the present application, the width of the sector area in the embodiment of the present application adopts a non-uniform width Width, taking into account the characteristics of the laser point cloud that is dense near and sparse far away.

[0033] For example, in a specific embodiment of the present application, the width of the near fan-shaped area is 2 m, the width of the middle fan-shaped area is 4 m, and the width of the far fan-shaped area is 8 m.

[0034] In other specific embodiments of the present application, the width of the near fan-shaped area is 4m, the width of the middle fan-shaped area is 8m, and the width of the far fan-shaped area is 12m.

[0035] It should be noted that the present application does not make any specific limitation on the width of the fan-shaped area, but rather gradually increases from near to far in the fan-shaped area segment, with the width of the near fan-shaped area being smaller than that of the middle fan-shaped area, and the width of the middle fan-shaped area being smaller than that of the far fan-shaped area.

[0036] Furthermore, in other embodiments of the present application, the division of the near area segment, the middle area segment, and the far area segment can be adaptively adjusted. For example, in a specific embodiment of the present application, 0-18m is a near sector segment, 18-36m is a middle sector segment, and 36 or more is a far sector segment, and the angle direction uniformly adopts 5° as an interval. In other specific embodiments of the present application, 0-15m can be used as a near sector segment, 15-30m is a middle sector segment, and 30 or more is a far sector segment, and the angle direction uniformly adopts 3° as an interval.

[0037] Finally, we get the grid array, traverse all the point clouds, and calculate the grid array according to each data point (R xy ,Theta) information to generate a coarse grid point cloud cluster P cell .

[0038] This application adopts a strategy of coarse detection followed by fine detection. Through coarse detection and fine detection, the target object is finally detected by combining the thickness and distance of the block point cloud and the horizontal gradient and other features. Among them, the coarse detection scheme is to use the maximum Z value point p zmax And the minimum Z value point p zmin Distance from the center point D zmax and D zmin , determine whether there is a convexity or concave in the area. If there is a convexity or concave in the grid area, further secondary segmentation is performed to implement a precise detection solution. For details, please refer to the embodiments later in this application.

[0039] Step S12: Obtain the thickness and center point distance of each coarse grid point cloud cluster.

[0040] Specifically, the present application proposes an embodiment, as a sub-step of step S12, for determining the center point distance. Figure 2 , Figure 2 The ground plane detection method provided by this application Figure 1 Schematic diagram of the sub-steps of step S12.

[0041] Step S121: Determine the thickness of the coarse grid point cloud cluster based on the difference between the maximum ordinate value and the minimum ordinate value in each coarse grid point cloud cluster.

[0042] Specifically, the ground plane detection device traverses each sector grid and calculates each grid point cloud cluster P cell_i The center point center(x, y, z) of the point cloud cluster is calculated. max And the minimum Z value Z min , and then calculate the thickness H of the point cloud cluster, H = (Z max -Z min ).

[0043] Step S122: Obtain the center point of the coarse grid point cloud cluster.

[0044] In a specific embodiment of the present application, the average coordinate values ​​of all points are calculated to obtain the coordinates of the center point of the point cloud. The PCL Point class can be used to represent a point in the point cloud. The geometric center point of the point cloud can be obtained by traversing all points and accumulating their coordinate values, and then dividing by the total number of points.

[0045] It should be noted that any method may be used to calculate the center point of the point cloud cluster, and in the embodiment of the present application, the specific method of obtaining the center point is not limited.

[0046] Step S123: determining the distance between the protrusion centers based on the difference between the maximum ordinate value and the ordinate value of the center point; determining the distance between the concave centers based on the difference between the minimum ordinate value and the ordinate value of the center point.

[0047] Specifically, the ordinate of the data point with the largest ordinate among all data points is subtracted from the ordinate of the center point, and the difference is used as the distance between the center point of the convex point cloud; the ordinate of the data point with the smallest ordinate among all data points is subtracted from the ordinate of the center point, and the difference is used as the distance between the center point of the concave point cloud.

[0048] Existing methods include solutions that integrate lidar and vision. LiDAR pothole detection obtains target obstacles through ground segmentation, downsampling, and clustering. The detection accuracy in existing technologies is not high because the ground segmentation algorithm is usually insensitive to potholes. Normal road surfaces have slight elevation changes, and the threshold needs to be adjusted slightly larger. In this way, potholes will be mistakenly segmented into the ground, and the target cannot be detected by clustering.

[0049] This application uses the camera semantic segmentation to extract the ground plane, and then uses the ground plane area of ​​the image to extract the ground plane of the point cloud, which can effectively ensure the accuracy of the point cloud ground segmentation and can also cope with steep slopes. Then, the ground plane block domain is refined by rasterizing the point cloud, and the ground plane is divided into small blocks to find ground defects, which can effectively ensure the accuracy of defect detection. The coarse detection is evaluated based on the thickness of the point cloud and the distance between high and low points to improve the accuracy of ground plane detection.

[0050] Step S13: Determine the concave-convex target coarse grid point cloud cluster according to the thickness and the center point distance.

[0051] The ground plane detection device determines the coarse grid point cloud cluster whose thickness is greater than the preset thickness threshold and the distance between the convex center points is greater than the preset convex threshold as the concave-convex target coarse grid point cloud cluster; and determines the coarse grid point cloud cluster whose thickness is greater than the preset thickness threshold and the distance between the concave center points is greater than the preset concave threshold as the concave-convex target coarse grid point cloud cluster.

[0052] Specifically, the ground plane detection device calculates the point p with the largest Z coordinate in the point cloud cluster zmax With the center point p center The distance D in the Z direction zmax and the largest point p zmin With the center point p center The distance D in the Z direction z,in , where D zmax =Z max -Z center ,D zmin =Z center -Z min The specific judgment is as follows: when H>H thr , D zmax >D thr There is a bulge in this area, H>H thr , D zmin >D thr , there is a depression in this area.

[0053] Step S14: Divide the concave-convex target coarse grid point cloud cluster into a number of fine grid point cloud clusters.

[0054] If there are convexities or concavities in the coarse grid area, the plane detection device further detects the coarse grid P cell_i The laser point cloud horizontal resolution is divided twice along the heading angle direction to obtain multiple small fan-shaped areas P. cell_ij , where the preset multiple can be any multiple such as 2 times, 3 times, 4 times, etc., and the specific multiple setting can be adaptively adjusted according to the required sector grid accuracy.

[0055] In other embodiments of the present application, other multiples may be used for division, and the present application does not make any specific limitation.

[0056] Step S15: Obtain the curvature of adjacent points of each fine grid point cloud cluster.

[0057] Before obtaining the curvature of adjacent points of each of the fine grid point cloud clusters, the ground plane detection method further includes: sorting the data points in the fine grid point cloud cluster according to horizontal distance.

[0058] Through secondary rasterization, the fan-shaped area P is obtained cell_ij , sort the small sector point cloud according to the horizontal distance, and then calculate the curvature Curva between adjacent data points of the sorted point cloud. The ground plane detection device calculates the small sector area P cell_ij The curvature curva along the horizontal distance direction is used to determine whether there is a target in the area based on the change in curvature.

[0059] Step S16: Obtain the concave and convex target point cloud in the fine grid point cloud cluster according to the curvature of adjacent points.

[0060] Specifically, in the embodiment of the present application, when the ground plane detection device determines that the fine grid point cloud cluster has a curvature increase or decrease according to the curvature of the adjacent points, the concave and convex target point cloud corresponding to the curvature change position is obtained. When the boundary point cloud of the fine grid point cloud cluster is determined to have a monotonically increasing or monotonically decreasing curvature according to the curvature of the adjacent points, the fine grid point cloud cluster is associated with the adjacent coarse grid point cloud cluster to obtain an associated point cloud cluster; when the associated point cloud cluster is determined to have a curvature increase or decrease according to the curvature of the adjacent points, the concave and convex target point cloud corresponding to the curvature change position is obtained. The sorted point cloud P cell_ij Calculate the curvature Curva between adjacent points. The calculation formula is as follows: C mn =(Z m -Z n ) / (X m -X n ).

[0061] When adjacent points are approximately on the same plane, the denominator is 0. If the denominator is not 0, it can be determined that there is a curvature change at a high position. When it is monotonically increasing or monotonically decreasing, the region is uphill or downhill, or there is monotony in a small section at the end, then it is necessary to associate with the adjacent large sector and record the coordinates of the point where the curvature changes; if the sector increases first and then decreases or decreases first and then increases, save the point where the curvature changes in the pothole target point cloud. In a sector, this change happens to be the starting position of the pothole bulge. Record the point clouds at these positions to obtain the target point cloud.

[0062] This application is based on the perception detection of laser radar, which makes full use of the laser radar's perception information of road surface elevation. It can perceive the information of road surface depressions as well as road surface protrusions, and identify the location of depression points through the change of elevation curvature.

[0063] Step S17: Output the concave-convex target on the ground plane using the concave-convex target point cloud.

[0064] For the target point cloud P obj The Andrew's convex polygon algorithm is used to generate the convex hull, and the maximum and minimum X, Y information is counted. Combined with the previous thickness information, the 3D boxing of the pothole or raised target is generated.

[0065] In one embodiment of the present application, a ground plane detection device obtains a ground plane point cloud image and a drivable area in an original image; maps pixels of the drivable area to the ground plane point cloud image and extracts the drivable area point cloud image; and extracts the ground plane point cloud from the original point cloud based on the drivable area point cloud image.

[0066] The ground plane detection device obtains an original point cloud; converts the original point cloud to a camera coordinate system to obtain a ground plane camera point cloud; and converts the camera point cloud to an image coordinate system to obtain the ground plane point cloud image.

[0067] Specifically, the present application obtains the coordinate transformation matrix M between the laser radar coordinate system and the camera coordinate system through the external parameter calibration algorithm, and transforms the original point cloud data P lidar Go to the camera coordinate system P camera , and at the same time record the index of each point in the principle point cloud to form an index vector V index . According to the camera internal parameters, the 3D point cloud P camera Projected to the physical coordinate system of the camera image, and finally transferred to the camera image pixel coordinate system to obtain the point cloud image I lidar .

[0068] Existing ground defect detection methods usually rely on visual prediction information about defects. This information is usually sufficient for large defects, but difficult for small ones. It is also difficult to detect potholes on curved surfaces that have been run over by wheels. Visual target detection requires that the two types of targets have obvious texture differences. However, the texture differences of small potholes and curved surfaces that have been run over by wheels are not very large, which will seriously affect the target detection performance.

[0069] This application can avoid such problems. Visual segmentation only selects the road surface without identifying whether there are bumps or depressions in the road surface. Visual segmentation has a relatively high segmentation accuracy for the road surface. Then, through information association, the road surface area in the laser point cloud is extracted, and the defects of the road surface are identified through a relatively high-precision laser sensor.

[0070] This application uses a two-step detection process of coarse detection and fine detection, and designs two fan-shaped rasterization methods to divide the ground plane into preset shapes. Then, the target object is finally detected by combining the thickness and distance of the block point cloud as well as the horizontal gradient and other features to improve the detection accuracy.

[0071] In order to implement the ground plane detection method of the above embodiment, the present application also provides a ground plane detection device. Figure 3 , Figure 3 It is a structural schematic diagram of an embodiment of a ground plane detection device provided in the present application.

[0072] like Figure 3 As shown, the ground plane detection device 600 of this embodiment includes a processor 61, a memory 62, an input and output device 63 and a bus 64.

[0073] The processor 61 , the memory 62 , and the input / output device 63 are respectively connected to the bus 64 . The memory 62 stores a computer program. The processor 61 is used to execute the computer program to implement the ground plane detection method of the above embodiment.

[0074] In this embodiment, the processor 61 may also be referred to as a CPU (Central Processing Unit). The processor 61 may be an integrated circuit chip having the ability to process signals. The processor 61 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The processor 61 may also be a GPU (Graphics Processing Unit), also known as a display core, a visual processor, or a display chip, which is a microprocessor that is specifically used for image computing on computers, workstations, game consoles, and some mobile devices (such as tablet computers, smart phones, etc.). The purpose of the GPU is to convert and drive the display information required by the computer system, and to provide a line scan signal to the display to control the correct display of the display. It is an important component that connects the display and the computer motherboard. As an important component of the computer host, the graphics card is responsible for outputting display graphics. The general-purpose processor may be a microprocessor or the processor 61 may also be any conventional processor, etc.

[0075] The present application also provides a computer storage medium, such as Figure 4 As shown, the computer storage medium 700 is used to store a computer program 71. When the computer program 71 is executed by a processor, it is used to implement the method described in the embodiment of the ground plane detection method of the present application.

[0076] The method involved in the embodiment of the ground plane detection method of the present application, when implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a device, such as a computer-readable storage medium. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (processor) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code.

[0077] The above description is only an implementation mode of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A ground plane detection method, characterized in that: The ground plane detection method comprises: Project the ground plane point cloud onto a preset sector grid to generate several coarse grid point cloud clusters; Obtaining the thickness and center point distance of each coarse grid point cloud cluster; Determine a concave-convex target coarse grid point cloud cluster according to the thickness and the center point distance; Dividing the concave-convex target coarse grid point cloud cluster into a plurality of fine grid point cloud clusters; Obtaining the curvature of adjacent points of each of the fine grid point cloud clusters; Acquire the concave-convex target point cloud in the fine grid point cloud cluster according to the adjacent point curvature; The concave-convex target point cloud is used to output the concave-convex target of the ground plane.

2. The ground plane detection method according to claim 1, characterized in that: The step of obtaining the thickness and center point distance of each coarse grid point cloud cluster includes: Determining the thickness of the coarse grid point cloud cluster based on the difference between the maximum ordinate value and the minimum ordinate value in each of the coarse grid point cloud clusters; Obtaining the center point of the coarse grid point cloud cluster; Determine the protrusion center point distance based on the difference between the maximum ordinate value and the ordinate value of the center point; Determine the distance between the center points of the depressions based on the difference between the minimum ordinate value and the ordinate value of the center point; The determining the concave-convex target coarse grid point cloud cluster according to the thickness and the center point distance comprises: Determine the coarse grid point cloud cluster whose thickness is greater than a preset thickness threshold and whose convex center point distance is greater than a preset convex threshold as the concave-convex target coarse grid point cloud cluster; And, the coarse grid point cloud cluster whose thickness is greater than a preset thickness threshold and whose concave center point distance is greater than a preset concave threshold is determined as the concave-convex target coarse grid point cloud cluster.

3. The ground plane detection method according to claim 1, characterized in that: Before obtaining the curvature of adjacent points of each of the fine grid point cloud clusters, the ground plane detection method further includes: The data points in the fine grid point cloud cluster are sorted according to horizontal distance.

4. The ground plane detection method according to claim 3, characterized in that: The step of acquiring the concave-convex target point cloud in the fine grid point cloud cluster according to the adjacent point curvature comprises: When it is determined according to the curvature of the adjacent points that the fine grid point cloud cluster has a curvature increase or decrease change, a concave-convex target point cloud corresponding to the curvature change position is obtained.

5. The ground plane detection method according to claim 3, characterized in that: The step of acquiring the concave-convex target point cloud in the fine grid point cloud cluster according to the adjacent point curvature comprises: When it is determined according to the adjacent point curvatures that the boundary point cloud of the fine grid point cloud cluster has a monotonically increasing or monotonically decreasing curvature, the fine grid point cloud cluster is associated with the adjacent coarse grid point cloud cluster to obtain an associated point cloud cluster; When it is determined according to the curvature of the adjacent points that the associated point cloud cluster has a curvature increase or decrease change, a concave-convex target point cloud corresponding to the curvature change position is obtained.

6. The ground plane detection method according to claim 1, characterized in that: The projecting of the ground plane point cloud onto a preset sector grid to generate a number of coarse grid point cloud clusters includes: Obtaining the horizontal distance and horizontal heading angle of the data points of the ground plane point cloud; The ground plane point cloud is projected onto a preset fan-shaped grid according to the horizontal distance of the data points and the horizontal heading angle to generate a plurality of coarse grid point cloud clusters.

7. The ground plane detection method according to claim 1, characterized in that: The ground plane detection method further includes: Obtain a ground plane point cloud image and a drivable area in the original image; Mapping pixels of the drivable area to the ground plane point cloud image to extract the drivable area point cloud image; Based on the drivable area point cloud image, the ground plane point cloud is extracted from the original point cloud.

8. The ground plane detection method according to claim 7, characterized in that: The step of obtaining a ground plane point cloud image comprises: Get the original point cloud; Convert the original point cloud to a camera coordinate system to obtain a ground plane camera point cloud; The camera point cloud is converted into an image coordinate system to obtain the ground plane point cloud image.

9. A ground plane detection device, characterized in that: The ground plane detection device includes a memory and a processor coupled to the memory; The memory is used to store program data, and the processor is used to execute the program data to implement the ground plane detection method according to any one of claims 1 to 7.

10. A computer storage medium, characterized in that: The computer storage medium is used to store program data, and when the program data is executed by a computer, it is used to implement the ground plane detection method according to any one of claims 1 to 7.

Images