Obstacle Recognition Method Based on Vehicle-mounted LiDAR

Abstract

The present invention discloses an obstacle recognition method based on an on-vehicle lidar, which relates to the technical field of point cloud data processing and includes: acquiring original point cloud data; performing preprocessing operations on the original point cloud data to obtain preprocessed point cloud data; performing ground segmentation on the preprocessed point cloud data to obtain obstacle point cloud data; constructing a polar coordinate grid map according to the retained obstacle point cloud data; wherein, the obstacle polar coordinate grid map includes a plurality of grids; obtaining an adaptive radial search step length for grid clustering according to a relative distance threshold; using the region growing method to label the grids that meet the requirements of the adaptive radial search step length to obtain different categories of grid sets; converting the different categories of grid sets into different categories of point cloud data sets to obtain obstacle information. The present invention can improve the recognition accuracy on the basis of ensuring the rapid recognition of obstacle targets.

Description

Technical Field

[0001] The present invention belongs to the technical field of point cloud data processing, and particularly relates to an obstacle recognition method based on an in-vehicle lidar. Background Art

[0002] In autonomous driving, the environmental perception system is the greatest guarantee for the driving safety of vehicles, providing basic support for high-level tasks such as subsequent vehicle path planning and behavior decision-making. Among them, the point cloud data based on lidar can provide the depth information of the target and has low sensitivity to weather and light. Its disadvantage is that the point cloud data has the characteristics of disorder, sparsity and unstructuredness, which increases the difficulty of lidar application.

[0003] In the prior art, lidar target detection methods can be divided into traditional methods and deep learning methods. Among them, the traditional lidar obstacle detection algorithm includes preprocessing, ground segmentation, obstacle clustering, etc. Among them, obstacle clustering is an important part of the traditional lidar target detection algorithm, which can enable the vehicle control system to fully identify information such as the shape, size, angle, and position of obstacles, and adopt effective obstacle avoidance strategies. In "ESTERMHPX.A Density-based Algorithm for Discovering Clusters a Density-based Algorithm for Discovering Clusters in Large Spatial Databases with Noise[J].International Conference on Knowledge Discovery and Data Mining, Portland, Oregon: AAAI, 1996: 226-231", a density-based clustering algorithm was proposed by introducing the density concept. This method can connect adjacent regions within the density threshold and can effectively handle outliers. Considering the characteristic that the lidar distance causes the point cloud to be dense near and sparse far, in "ZHENG L, ZHANG P, TAN J, et al.The Obstacle Detection Method of UAV Based on 2D Lidar[J].IEEE ACCESS, 2019, 7: 163437-163448", a clustering algorithm based on relative distance and density is adopted. It not only corrects the offset of the point cloud but also effectively clusters the point cloud with uneven density. In "XUE P, WU Y, YIN G, et al.Real-time Target Recognition for Urban Autonomous Vehicles Based on Information Fusion[J].In "Jixie GongchengXuebao / Journal of Mechanical Engineering, 2020, 56(12): 165 - 173", it is proposed to generate an adaptive radius at this position by combining the threshold radius and the point cloud distance, which improves the robustness of the algorithm compared with the traditional fixed radius; the density - based clustering algorithm has a good clustering effect, but the time - consuming problem is very serious, and the grid - based clustering algorithm can solve this problem; the main idea of the grid - based clustering algorithm is to first quantize the space into a finite number of cells, and compress and represent the data according to the statistical information within the cells. Due to the scattered distribution of the point cloud, the distance between the point clouds in the distance is relatively far, resulting in the over - segmentation of distant obstacles by such methods; aiming at the shortcomings of traditional grid - based clustering, Himmelsbach M, Mueller A, Lüttel T, et al. proposed creating a polar coordinate grid coordinate system for clustering in "LIDAR - based 3D object perception[C] / / Proceedings of 1st international workshop on cognition for technical systems. 2008, 1", which improves the clustering accuracy to a certain extent. Although the grid - based clustering algorithm has good real - time performance, the problem of being unable to process irregularly distributed data still exists.

[0004] Therefore, it is urgent to improve the above - mentioned problems existing in the prior art. Summary of the Invention

[0005] In order to solve the above - mentioned problems existing in the prior art, the present invention provides an obstacle recognition method based on a vehicle - mounted lidar. The technical problems to be solved by the present invention are realized through the following technical solutions:

[0006] In the first aspect, the present invention provides an obstacle recognition method based on a vehicle - mounted lidar, including:

[0007] Obtain the original point cloud data;

[0008] Perform pre - processing operations on the original point cloud data to obtain pre - processed point cloud data; then perform ground segmentation on the pre - processed point cloud data to obtain obstacle - retained point cloud data;

[0009] Construct a polar coordinate grid map according to the obstacle - retained point cloud data; among them, the obstacle polar coordinate grid map includes a plurality of grids;

[0010] Obtain the adaptive radial search step of grid - based clustering according to the relative distance threshold;

[0011] Use the region growing method to label the grids that meet the requirements of the adaptive radial search step, and obtain grid sets of different categories;

[0012] Convert the grid sets of different categories into point cloud data sets of different categories to obtain obstacle information.

[0013] Advantages of the present invention:

[0014] An obstacle recognition method based on a vehicle-mounted lidar provided by the present invention performs preprocessing and ground segmentation operations on the original point cloud data to obtain the remaining obstacle point cloud data; constructs a polar coordinate grid map based on the remaining obstacle point cloud data; by traversing the grids in the polar coordinate grid map, while keeping the relative distance threshold unchanged, calculates the adaptive radial search step of the current grid; and uses this grid as the central grid to search for grids that meet the requirements of the adaptive radial search step to obtain grid sets of different categories; then converts the grid sets of different categories into point cloud data sets of different categories to obtain obstacle information; in this way, aiming at the problem that irregularly distributed data cannot be processed in grid clustering in the prior art, the present invention proposes a relative distance calculation method for two grids, reduces the time complexity of calculating the relative distance between point clouds, and obtains the conversion relationship between two-dimensional coordinate points and grids while keeping the relative distance threshold unchanged, obtains the adaptive radial search step, that is, the step size is small in the near range and large in the far range, and solves the problem brought by the uneven density of point cloud data near and far while ensuring real-time performance, and can improve the recognition accuracy on the basis of ensuring the rapid recognition of obstacle targets.

[0015] The following will further describe the present invention in detail with reference to the drawings and embodiments. Description of the Drawings

[0016] Figure 1 is a flowchart of an obstacle recognition method based on a vehicle-mounted lidar provided by an embodiment of the present invention;

[0017] Figure 2 is a schematic diagram of the ground point cloud data being circular after visualization provided by an embodiment of the present invention;

[0018] Figure 3 is a schematic diagram of the point cloud data after the ground segmentation operation provided by an embodiment of the present invention;

[0019] Figure 4 is a schematic diagram of a grid polar coordinate map provided by an embodiment of the present invention;

[0020] Figure 5 is a schematic diagram of obtaining an obstacle provided by an embodiment of the present invention;

[0021] Figure 6 is a schematic diagram of obstacle conversion provided by an embodiment of the present invention;

[0022] Figure 7 It is a flowchart for obtaining different types of grid sets provided by an embodiment of the present invention;

[0023] Figure 8 It is a schematic diagram for obtaining obstacles provided by an embodiment of the present invention;

[0024] Figure 9 It is a schematic diagram for the clustering effect provided by an embodiment of the present invention. Detailed implementation manners

[0025] The following further describes the present invention in detail with reference to specific embodiments, but the implementation manners of the present invention are not limited thereto.

[0026] In the prior art, grid clustering is a common clustering method. Taking each grid as the search unit, it has a lower time complexity. The most important advantage of this method is good real-time performance. However, the improvement effect of the existing grid clustering on clustering by only adjusting the grid size is limited and cannot handle complex scenarios; based on the density clustering algorithm, a more accurate clustering result can be obtained, but the time complexity of this method is high and does not meet the real-time requirements.

[0027] In view of this, the present invention provides an obstacle recognition method based on a vehicle-mounted lidar, which can solve the problems of false detection and missed detection in vehicle-mounted lidar obstacle recognition due to the characteristics of point cloud data being dense near and sparse far in complex scenarios and scenarios with changing environments, and can improve the recognition accuracy on the basis of ensuring the rapid recognition of obstacle targets.

[0028] Please refer to Figure 1 as shown in Figure 1 It is a flowchart for an obstacle recognition method based on a vehicle-mounted lidar provided by an embodiment of the present invention. An obstacle recognition method based on a vehicle-mounted lidar provided by the present invention includes:

[0029] S101. Obtain original point cloud data;

[0030] S102. Perform preprocessing operations on the original point cloud data to obtain preprocessed point cloud data; then perform ground segmentation on the preprocessed point cloud data to obtain obstacle point cloud data retained;

[0031] S103. Construct a polar coordinate grid map according to the obstacle point cloud data retained; wherein, the obstacle polar coordinate grid map includes a plurality of grids;

[0032] S104. Obtain an adaptive radial search step size for grid clustering according to a relative distance threshold;

[0033] S105. Use the region growing method to label the grids that meet the requirements of the adaptive radial search step, and obtain grid sets of different categories;

[0034] S106. Convert the grid sets of different categories into point cloud data sets of different categories to obtain obstacle information.

[0035] Specifically, please continue to refer to Figure 1 As shown, an obstacle recognition method based on vehicle-mounted lidar provided in this embodiment preprocesses the original point cloud data and performs ground segmentation operations to obtain the remaining obstacle point cloud data; constructs a polar coordinate grid map based on the remaining obstacle point cloud data; by traversing the grids in the polar coordinate grid map, while keeping the relative distance threshold unchanged, calculates the adaptive radial search step of the current grid; and uses this grid as the central grid to search for grids that meet the requirements of the adaptive radial search step to obtain grid sets of different categories; then converts the grid sets of different categories into point cloud data sets of different categories to obtain obstacle information; in this way, aiming at the problem that irregularly distributed data cannot be processed in grid clustering in the prior art, the present invention proposes a relative distance calculation method for two grids, reduces the time complexity of calculating the relative distance between point clouds, obtains the conversion relationship between two-dimensional coordinate points and grids while keeping the relative distance threshold unchanged, obtains the adaptive radial search step, that is, the step size is small near and large far away, and solves the problem brought by the uneven density of point cloud data near and far while ensuring real-time performance, and can improve the recognition accuracy on the basis of ensuring the rapid recognition of obstacle targets.

[0036] In an optional embodiment of the present invention, the process of preprocessing the original point cloud data to obtain the preprocessed point cloud data includes:

[0037] Use the road edge detection algorithm to perform through filtering on the original point cloud data to obtain an effective point cloud area;

[0038] Perform denoising processing on the original point cloud data after through filtering processing to obtain the processed point cloud data.

[0039] Specifically, in this embodiment, the preprocessing operation on the original point cloud data is filtering processing. The filtering processing is applicable to the vehicle driving road scenario. The through filtering processing is performed on the point cloud data on both sides of the road by using the road edge detection algorithm, which can reduce the number of irrelevant point clouds; after the through filtering processing, an effective point cloud area is obtained. In addition, due to the influence of equipment accuracy and actual environment in the process of collecting point cloud data, the original point cloud data often contains noise points, and it is necessary to perform denoising processing on the point cloud data.

[0040] In an optional embodiment of the present invention, use the ground segmentation algorithm based on plane fitting to perform ground segmentation on the preprocessed point cloud data.

[0041] Specifically, please refer to Figure 2 and Figure 3 as shown. Figure 2 Fig. is a schematic diagram of the visualized ground point cloud data being circular provided by an embodiment of the present invention. Figure 3 Fig. is a schematic diagram of the point cloud data after the ground segmentation operation provided by an embodiment of the present invention. In this embodiment, in the point cloud data, the ground points are the points reflected by the lidar scanning on the ground. Please continue to refer to Figure 2 as shown. Since the visualized ground point cloud data is circular, it may cause the connection between the ground and the targets above the ground, which is not conducive to the subsequent clustering operation. Therefore, it is necessary to perform segmentation processing on the ground point cloud and only retain the obstacle point cloud data. Optionally, in this embodiment, a ground segmentation algorithm based on plane fitting is used to filter out the ground, and the result of retaining the obstacle point cloud data is as shown in Figure 3 as shown.

[0042] In an alternative embodiment of the present invention, please refer to Figure 4 as shown. Figure 4 Fig. is a schematic diagram of a grid polar coordinate map provided by an embodiment of the present invention. The process of constructing a polar coordinate grid map based on the retained obstacle point cloud data includes:

[0043] Converting the retained obstacle point cloud data into two-dimensional coordinate data;

[0044] Converting the two-dimensional coordinate data from Cartesian rectangular coordinate data to grid polar coordinate data to obtain a grid polar coordinate map.

[0045] Specifically, please continue to refer to Figure 4 as shown. Considering that the three-dimensional information has a high calculation complexity and long time consumption, which is not conducive to the real-time application of the algorithm, in this embodiment, the retained obstacle point cloud data is projected onto the XOY plane, that is, the three-dimensional information of the point cloud data is converted into two-dimensional coordinate data, which is conducive to subsequent processing. The two-dimensional coordinate data projected onto the XOY plane, that is, point i, with coordinates p i (x i , y i ), is converted from Cartesian rectangular coordinate data to grid polar coordinate data to obtain an M Bins ×N Channels grid polar coordinate map. Among them, the conversion expression is:

[0046]

[0047]

[0048] Among them, Channel(p i ) and Bin(p i) is the point p i (x i , y i ) is the converted grid polar coordinate, is the floor function, α is the angular interval of the grid coordinate system, and R is the radius interval of the grid coordinate system.

[0049] In an alternative embodiment of the present invention, before converting the two-dimensional coordinate data into grid polar coordinate data, it further includes:

[0050] Taking the lidar center as the pole, the front of the lidar as the polar axis, and the counterclockwise direction as positive for the angle, a polar coordinate system is constructed;

[0051] According to the radius of the grid spacing between adjacent polar coordinates and the angle of the grid spacing between adjacent polar coordinates, with the lidar as the center and a 360° horizontal viewing angle range, a grid polar coordinate system is constructed; wherein, the grid polar coordinate system includes M rows and N columns of grids, and the expressions of M and N are respectively:

[0052]

[0053]

[0054] wherein, VD is the visible distance of the lidar model.

[0055] In an alternative embodiment of the present invention, the process of obtaining the adaptive radial search step of grid clustering according to the relative distance threshold includes:

[0056] Obtain the relative distance threshold;

[0057] Obtain the radius coordinate of the center point j of the grid that is at a distance of the relative distance threshold from the center point i of the current grid, and its expression is:

[0058] r j = d × ρ' i + ρ' i ;

[0059] wherein, r j is the radius coordinate of the center point j of the grid, d is the relative distance threshold, and ρ' i is the radius coordinate of the center point i of the current grid; it should be noted that due to the characteristic of the lidar being denser near and sparser far away, the center point j of the grid is outside the center point i of the current grid, and ρ' B - ρ' A > 0;

[0060] According to the radius coordinate of the center point j of the grid, obtain the coordinate of the center point j of the grid, and according to the coordinate of the center point j of the grid, obtain the adaptive radial search step, and its expression is:

[0061]

[0062] Specifically, in this embodiment, the central coordinates of another grid are calculated through a given relative distance threshold and the polar coordinate position of one of the grids, and an adaptive radial search step size is further obtained. It should be noted that the concept of the relative distance threshold refers to the following calculation method of the relative distance, specifically:

[0063] Please refer to Figure 5 and Figure 6 as shown, Figure 5 which is a schematic diagram of obtaining an obstacle provided by an embodiment of the present invention, Figure 6 which is a schematic diagram of obstacle conversion provided by an embodiment of the present invention. Point O is the center point of the lidar, points A and B are the reflection points on the surface of the obstacle encountered by the lidar, and the calculation of the relative distance is similar to the scaling transformation of a triangle, as Figure 6 shown. Connect points A and B to construct triangle ABO, then shorten the edge OA to the unit length, and at the same time shorten the side AB and OB by the same ratio. With point A as the center, the expression of the relative distance between A and B is:

[0064]

[0065] where l AB is the distance between point A and point B, and l A is the distance from the lidar center to point A.

[0066] In the Cartesian rectangular coordinate system, the coordinates of point A are (x A , y A ), the coordinates of point B are (x B , y B ), and with point A as the midpoint, the expression of the relative distance between point A and point B is:

[0067]

[0068] In the polar coordinate system, when the angles of points A and B are the same, the polar coordinates of point A are (θ A , r A ), the polar coordinates of point B are (θ B , r B ), and with point A as the center, the expression of the relative distance between point A and point B is:

[0069]

[0070] In the grid polar coordinate system, the center of the grid represents the position information of the grid. To obtain the center point of the grid, the center points of grid A (θ' A , ρ'A ) and the center point of grid B (θ' B , ρ' B ); wherein, grid A and grid B are at the same angle;

[0071]

[0072]

[0073] Wherein, R is the radius of the adjacent grid spacing in the radial direction of the polar coordinate, α is the angle of the adjacent grid spacing in the polar coordinate, (θ A , r A ) is the polar coordinate of point A, and (θ B , r B ) is the polar coordinate of point B;

[0074] Taking grid A as the center, obtain the relative distance d between grid A and grid B AB , and its expression is:

[0075]

[0076] Above, by obtaining the relative distance, obtain the relative distance threshold required in this embodiment.

[0077] In an alternative embodiment of the present invention, please refer to Figure 7 shown, Figure 7 is a flowchart for obtaining a set of grids of different categories provided by an embodiment of the present invention. The process of using the region growing method to mark the grids that meet the requirements of the adaptive radial search step size and obtaining a set of grids of different categories includes:

[0078] Taking the current grid as the central grid, search all the grids in the connected region ε of the central grid, and sequentially determine whether the searched grid is the central grid; if it is the central grid, search all the grids in the connected region ε of the central grid; if it is not the central grid, stop the search, obtain a set of grids of one category, and assign the same number to the grids in the set of grids of one category; repeat the loop search, and according to the region growing rule, search all the grids to obtain a set of grids of different categories, and assign different numbers to the grids in the set of grids of different categories, record the maximum value of the numbers, which is the total number of obstacles in the current frame data; during the traversal process, if there is no point cloud in the grid, directly skip this grid; wherein, the central grid is a grid with the number of point clouds in the connected region ε greater than or equal to the minimum density MinPts.

[0079] Specifically, in this embodiment, it is necessary to continuously determine whether the currently searched grid is the central grid. According to the definition of the central grid, first, the range of the connected region needs to be determined, and then the number of point clouds in the connected region is calculated and compared with the MinPts threshold to determine the grid attribute. With the development of the density clustering algorithm, the adaptive neighborhood range improves the clustering accuracy, and a similar conclusion can also be obtained in grid clustering. Therefore, when determining the neighborhood range in grid clustering, the calculation of the adaptive threshold can effectively improve the clustering accuracy.

[0080] Generally, to determine whether two points belong to the same cluster, the distance between the two points needs to be calculated and compared with the threshold. If it is greater than or equal to the threshold, the two points belong to the same cluster; if it is less than the threshold, the two points do not belong to the same cluster. Due to the characteristic of point cloud data that it is dense near and sparse far away, the distance threshold needs to change in real time to be used for detection in a larger range. The present invention designs the conversion relationship between points and grids by using the method of not changing the relative distance threshold, calculates the radial region search range of the currently traversed grid, so that the same threshold can be applied to each grid, reducing the changing parameters.

[0081] It should be noted that, please refer to Figure 8 as shown in Figure 8 is a schematic diagram of obtaining obstacles provided by an embodiment of the present invention. The connected region ε is all the grids within the grids centered on the current grid, with a connected angle of AngStep grids and a radial RStep grids. Search the grids, and the gray part is the grid connected region, that is, the region where AngStep is 1 and RStep is 1; the central grid is the grid with the number of point clouds in the connected region ε greater than or equal to the minimum density MinPts.

[0082] In an optional embodiment of the present invention, the process of converting the grid sets of different categories into point cloud data sets of different categories to obtain obstacle information includes:

[0083] Convert the polar coordinate data in the grid set into Cartesian rectangular coordinate data, then convert the Cartesian rectangular coordinate data into two-dimensional coordinate data, and further convert it into point cloud data to obtain an obstacle information.

[0084] Convert the polar coordinate data in different grid sets to obtain multiple obstacle information.

[0085] Specifically, in this embodiment, according to the requirements of on-vehicle lidar obstacle recognition, it is not enough to only complete the grid class marking. It is also necessary to convert the grid set obtained in the above process. Each grid set of a class is abstracted into an obstacle set, and the information included in this obstacle set is shown in Table 1.

[0086] Table 1 Information included in the obstacle set

[0087] Information Name Data Form Number ID Point Cloud Data P(x,y,z,i) Center Position O(x,y,z) Boundary Contour <![CDATA[BBox{b1,…,b n}]]> Length, Width, Height Length,Width,Height

[0088] Create an obstacle set. The number of objects in the set is the number of clusters in the above process, that is, the number of identified obstacles, and assign a unique number to each object in the current frame. This number corresponds one-to-one with the grid category. Traverse the point cloud data, and according to the Cartesian rectangular coordinate system and polar coordinate grid mapping relationship in the formula, assign the same label of the corresponding grid to each point cloud, and store the point cloud into the object with the corresponding number according to the relationship between the cluster grid label and the set object number.

[0089] In an optional embodiment of the present invention, the obstacle information includes the boundary contour of the obstacle, the length, width and height of the obstacle, and the center of the obstacle. Among them, the boundary contour of the obstacle is the vertex information composed of the maximum and minimum values of the x-axis, y-axis and z-axis of the point cloud, the length, width and height of the obstacle are the length information of the difference between the maximum and minimum values of the x-axis, y-axis and z-axis of the point cloud, and the center of the obstacle is the geometric center of the obstacle boundary.

[0090] In an optional embodiment of the present invention, please refer to Figure 9 as shown in Figure 9 which is a schematic diagram of the clustering effect provided by the embodiment of the present invention. Compared with the traditional grid clustering algorithm and density clustering algorithm, the method provided in this embodiment has higher clustering accuracy. Figure 9 It shows that the accuracy within 20 meters, 30 meters, and 50 meters of the adaptive search step is higher than that of the traditional grid clustering algorithm and the traditional density clustering algorithm, and the total error rate is significantly reduced. Please refer to Table 2 as shown.

[0091] Table 2 Clustering Accuracy

[0092]

[0093]

[0094] It should be noted that in this text, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant are intended to cover non-exclusive inclusion, so that an article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the article or device comprising said element. Words such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The orientation or positional relationship indicated by "above", "below", "left", "right", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0095] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. means that the specific features or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not have to be directed to the same embodiment or example. Moreover, the specific features or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine the different embodiments or examples described in this specification.

[0096] The above content is a further detailed description of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is only limited to these descriptions. For those of ordinary skill in the technical field to which the present invention pertains, without departing from the concept of the present invention, several simple deductions or substitutions can still be made, and all should be regarded as belonging to the protection scope of the present invention.

Claims

1. An obstacle recognition method based on vehicle-mounted lidar, characterized in that, Including: Obtaining original point cloud data; Performing preprocessing operations on the original point cloud data to obtain preprocessed point cloud data; Performing ground segmentation on the preprocessed point cloud data to obtain obstacle point cloud data; Constructing a polar coordinate grid map based on the obstacle point cloud data; wherein, the obstacle polar coordinate grid map includes a plurality of grids; Obtaining an adaptive radial search step size for grid clustering according to a relative distance threshold, including: Obtaining the relative distance threshold; Obtain the center point of the current grid The center points of the grids that are at the relative distance threshold from it The radius coordinates, and its expression is: ； Among them, is the center point of the grid is the radius coordinate of is the relative distance threshold, is the center point of the current grid is the radius coordinate; According to the central point of the grid of the radius coordinates, obtain the coordinates of the central point of the grid of the coordinates, according to the central point of the grid of the coordinates, obtain the adaptive radial search step size, and its expression is: ； Among them, Grid coordinate system radius interval; Using the region growing method to label grids that meet the requirements of the adaptive radial search step size to obtain grid sets of different categories; Converting the grid sets of different categories into point cloud data sets of different categories to obtain obstacle information.

2. The obstacle recognition method based on vehicle-mounted lidar according to claim 1, characterized in that, The process of performing preprocessing operations on the original point cloud data to obtain preprocessed point cloud data includes: Performing direct pass filtering on the original point cloud data using a road edge detection algorithm to obtain an effective point cloud region; Performing denoising processing on the original point cloud data after direct pass filtering to obtain processed point cloud data.

3. The obstacle recognition method based on an in-vehicle lidar according to claim 1, wherein Performing ground segmentation on the preprocessed point cloud data using a ground segmentation algorithm based on plane fitting.

4. The obstacle recognition method based on vehicle-mounted lidar according to claim 1, characterized in that, The process of constructing a polar coordinate grid map based on the obstacle point cloud data includes: Converting the obstacle point cloud data into two-dimensional coordinate data; Converting the two-dimensional coordinate data from Cartesian rectangular coordinate data into grid polar coordinate data to obtain a grid polar coordinate map.

5. The obstacle recognition method based on an in-vehicle lidar according to claim 4, wherein Before converting the two-dimensional coordinate data into grid polar coordinate data, it further includes: Constructing a polar coordinate system with the center of the lidar as the pole, the front of the lidar as the polar axis, and the counterclockwise direction as the positive direction for the angle; Construct a grid polar coordinate system centered on the lidar according to the radius of the polar coordinate radial adjacent grid spacing and the angle of the polar coordinate adjacent grid spacing within a 360° horizontal viewing angle range; wherein, the grid polar coordinate system includes rows, column grids.

6. The obstacle recognition method based on vehicle-mounted lidar according to claim 1, wherein The process of using the region growing method to label grids that meet the requirements of the adaptive radial search step size to obtain grid sets of different categories includes: Taking the current grid as the central grid, search for the connected area of the central grid for all grids within it, and sequentially determine whether the searched grid is the central grid; if it is the central grid, search for the connected area of this central grid for all grids within it; if it is not the central grid, stop the search, obtain a set of grids of one class, and assign the same number to the grids in the set of grids of one class; repeat the loop search, according to the region growing rule, search all grids, obtain sets of grids of different classes, and assign different numbers to the grids in the sets of grids of different classes; where the central grid is a grid within the connected area with the number of point clouds greater than or equal to the minimum density MinPts.

7. The obstacle recognition method based on the vehicle-mounted lidar according to claim 1, wherein The process of converting the grid sets of different categories into point cloud data sets of different categories to obtain obstacle information includes: Converting the polar coordinate data in the grid set into Cartesian rectangular coordinate data, then converting the Cartesian rectangular coordinate data into two-dimensional coordinate data, and further converting it into point cloud data to obtain one piece of obstacle information; Converting the polar coordinate data in different categories of grid sets to obtain multiple pieces of obstacle information.

8. The obstacle recognition method based on vehicle-mounted lidar according to claim 7, wherein, The obstacle information includes the boundary contour of the obstacle, the length, width and height of the obstacle, and the center of the obstacle; wherein, the boundary contour of the obstacle is point cloud axis, axis and the vertex information composed of the maximum and minimum values of the axis, and the length, width and height of the obstacle are point cloud axis, axis and the length information of the difference between the maximum and minimum values of the axis, and the center of the obstacle is the geometric center of the obstacle boundary.

Images