Negative obstacle area identification method and device and storage medium

Abstract

The invention provides a negative obstacle area identification method and device and a storage medium, and the method comprises the steps: carrying out the rasterization of point cloud data according to the obtained reference distance between the point cloud data and a vehicle, and generating a multi-size raster map; determining a grid label of each grid in the multi-size grid map according to the counted point cloud distribution characteristics; determining key feature point pairs of the to-be-identified grids according to the grid labels and set feature point pair identification conditions; and carrying out clustering analysis on the key feature point pairs, and identifying a negative obstacle region. According to the embodiment of the invention, the detection time of the negative obstacle on the unstructured road is reduced, and the detection efficiency is improved. Meanwhile, the negative obstacle area is detected by using various characteristics, the false detection probability and the missed detection probability are reduced, and the accuracy and the reliability of negative obstacle detection are improved, so that the vehicle can more safely adapt to complex road conditions with negative obstacles, and the driving safety of the vehicle is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of negative obstacle detection, and in particular to a negative obstacle area recognition method, device and storage medium. Background Art

[0002] On unstructured roads, the road surface is undulating and there are a large number of irregular negative obstacles. Negative obstacles refer to terrain that is lower than the surrounding ground, such as craters, puddles, trenches, and cliffs. In related technologies, the absolute coordinate plane method is usually used to identify negative obstacles, which has a high missed detection rate and false detection rate, resulting in low driving safety of vehicles. Summary of the invention

[0003] In view of the above problems, the embodiments of the present invention provide a negative obstacle area identification method, device and storage medium to solve the problem in the prior art that the negative obstacle missed detection rate and false detection rate are high, resulting in low driving safety of the vehicle.

[0004] In a first aspect, an embodiment of the present invention provides a method for identifying a negative obstacle area, the method comprising:

[0005] rasterizing the point cloud data according to a reference distance between the acquired point cloud data and the vehicle to generate a multi-size raster map;

[0006] Determine a grid label for each grid in the multi-size grid map according to the statistically determined point cloud distribution characteristics;

[0007] Determine the key feature point pairs of the grid to be identified according to the grid labels and the set feature point pair identification conditions;

[0008] Cluster analysis is performed on the key feature point pairs to identify negative obstacle areas.

[0009] In a possible implementation, the step of rasterizing the point cloud data according to the reference distance between the acquired point cloud data and the vehicle to generate a multi-size grid map includes:

[0010] According to the correspondence between the set reference distance and the grid size, the grid size corresponding to the reference distance is queried, and the point cloud data is divided into grids of the grid size until all point cloud data are traversed to generate a multi-size grid map.

[0011] In a possible implementation, the point cloud distribution features include a maximum height, a minimum height, and the number of point clouds.

[0012] In a possible implementation, the grid label includes a ground label, a negative obstacle label, a positive obstacle label, or an unknown label;

[0013] Determining the grid label of each grid in the multi-size grid map according to the statistical point cloud distribution characteristics includes:

[0014] Determine whether the number of point clouds in the grid is zero;

[0015] If it is determined that the number of point clouds in the grid is zero, the grid label of the grid is marked as an unknown label;

[0016] If it is determined that the number of point clouds in the grid is not zero, then determining whether a height difference is greater than a preset threshold, the height difference being used to indicate a difference between the maximum height and the minimum height;

[0017] If it is determined that the height difference is less than or equal to the preset threshold, marking the grid label of the grid as a ground label;

[0018] If it is determined that the height difference is greater than a preset threshold, determining whether the minimum height value is less than zero;

[0019] If it is determined that the minimum height value is less than zero, the grid label of the grid is marked as a negative obstacle label;

[0020] If it is determined that the minimum height value is greater than or equal to zero, the grid label of the grid is marked as a positive obstacle label.

[0021] In a possible implementation, the key feature point pair includes a first feature point and a second feature point;

[0022] The characteristic point pair recognition condition includes that the horizontal distance between points is greater than the horizontal distance between first adjacent points, and the horizontal distance between points is greater than the horizontal distance between second adjacent points;

[0023] The horizontal distance between points is used to indicate the horizontal distance between the first feature point and the second feature point, the horizontal distance between the first adjacent points is used to indicate the horizontal distance between the first feature point and the adjacent points of the first feature point, and the horizontal distance between the second adjacent points is used to indicate the horizontal distance between the second feature point and the adjacent points of the second feature point.

[0024] In a possible implementation manner, the feature point pair recognition condition further includes a height difference between the first feature point and the second feature point.

[0025] In a possible implementation manner, the feature point pair recognition condition further includes that a point cloud density at a location where the first feature point is located is different from a point cloud density at a location where the second feature point is located.

[0026] In a possible implementation, performing cluster analysis on the key feature point pairs to identify negative obstacle areas includes:

[0027] Clustering the central points of the key feature point pairs to generate at least one target point cluster;

[0028] Determine whether the number of center points of the target point cluster is greater than a preset number threshold;

[0029] If it is determined that the number of center points of the target point cluster is greater than or equal to the preset number threshold, the target point cluster is marked as a negative obstacle point cluster;

[0030] If it is determined that the number of center points of the target point cluster is less than the preset number threshold, marking the target point cluster as a ground point cluster;

[0031] A negative obstacle area is generated according to the at least one negative obstacle point cluster.

[0032] In a possible implementation, the target point cluster includes an initial point cluster and a newly created point cluster;

[0033] The step of clustering the central points of the key feature point pairs to generate at least one target point cluster includes:

[0034] Randomly select the center point of a key feature point pair as an initial node, and generate an initial point cluster with the initial node as the center;

[0035] Determine whether the distance between the center point to be allocated and the initial point cluster is less than a preset distance threshold, the center point to be allocated is used to indicate a center point outside the initial node;

[0036] If it is determined that the distance between the center point to be allocated and the initial point cluster is less than the preset distance threshold, the center point to be allocated is added to the initial point cluster;

[0037] If it is determined that the distance between the center point to be assigned and the initial point cluster is greater than or equal to the preset distance threshold, a new point cluster is generated with the center point to be assigned as the center until the center points of all key feature point pairs are traversed, so that each center point to be assigned is added to the target point cluster.

[0038] In a second aspect, an embodiment of the present invention provides a negative obstacle area identification device, the device comprising:

[0039] A generation module, used for rasterizing the point cloud data according to the reference distance between the acquired point cloud data and the vehicle to generate a multi-size raster map;

[0040] A label determination module, used to determine the grid label of each grid in the multi-size grid map according to the statistical point cloud distribution characteristics;

[0041] A determination module, used to determine the key feature point pairs of the grid to be identified according to the grid labels and the set feature point pair identification conditions;

[0042] The identification module is used to perform cluster analysis on the key feature point pairs to identify negative obstacle areas.

[0043] The technical solution provided by the embodiment of the present invention reduces the detection time of negative obstacles on unstructured roads, improves the detection efficiency, and has high real-time performance. At the same time, multiple features are used to detect negative obstacle areas, which reduces the probability of false detection and missed detection, and improves the accuracy and reliability of negative obstacle detection, so that the vehicle can more safely adapt to complex road conditions with negative obstacles, thereby improving the driving safety of the vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 FIG. 4 is a flow chart of a negative obstacle area identification method provided by an embodiment of the present invention.

[0045] Figure 2 Shown is a schematic diagram of a distance range provided by an embodiment of the present invention.

[0046] Figure 3 FIG. 1 is a schematic diagram of a key feature point pair recognition principle provided by an embodiment of the present invention.

[0047] Figure 4 A flowchart of a grid label determination method provided by an embodiment of the present invention.

[0048] Figure 5 A schematic diagram of the structure of a negative obstacle area identification device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present invention clearer and more understandable, the present invention is further described below in conjunction with the accompanying drawings and specific implementation methods. 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.

[0050] Figure 1 A flowchart of a negative obstacle area identification method provided by an embodiment of the present invention is as follows: Figure 1 As shown, the method includes:

[0051] Step 101: rasterize the point cloud data according to the reference distance between the acquired point cloud data and the vehicle to generate a multi-size raster map.

[0052] Each step in the embodiment of the present invention is executed by a domain controller.

[0053] In this step, the domain controller queries the grid size corresponding to the reference distance according to the correspondence between the set reference distance and the grid size, and divides the point cloud data into grids of the grid size until all point cloud data are traversed to generate a multi-size grid map. Specifically, the domain controller projects the point cloud data into a two-dimensional vehicle body coordinate system to generate a multi-size grid map. Among them, the two-dimensional vehicle body coordinate system is a two-dimensional coordinate system established with the center of the vehicle as the origin, the vertical direction of the vehicle's driving direction as the horizontal axis, and the vehicle's driving direction as the positive direction of the vertical axis.

[0054] In an embodiment of the present invention, a multi-size grid map has at least two grid sizes. In practical applications, the type of grid size can be set based on specific circumstances. In an embodiment of the present invention, three grid sizes are set as an example for description. For example, the three grid sizes are set to 10cm×10cm, 25cm×25cm, and 50cm×50cm, respectively. The correspondence between the reference distance and the grid size is set as follows: within the close range of the center of the vehicle, a 10cm×10cm grid is used to generate a grid map; within the medium range of the center of the vehicle, a 25cm×25cm grid is used to generate a grid map; within the long range of the center of the vehicle, a 50cm×50cm grid is used to generate a grid map.

[0055] Figure 2 A schematic diagram of a distance range provided by an embodiment of the present invention, such as Figure 2 As shown, the short-distance range can be set to a range extending from the origin along the positive and negative directions of the horizontal axis by five meters each, and along the positive direction of the vertical axis by ten meters, that is, area ①; the long-distance range can be set to a range outside the range extending from the origin along the positive and negative directions of the horizontal axis by ten meters each, and along the positive direction of the vertical axis by thirty meters, that is, area ③; the medium-distance range can be set to a range between the short-distance range and the long-distance range, that is, area ②.

[0056] In the embodiment of the present invention, the ground road condition close to the center of the vehicle has a greater impact on the driving safety of the vehicle, while the ground road condition far from the center of the vehicle has a smaller impact on the driving safety of the vehicle. Therefore, the road condition close to the center of the vehicle is projected into a grid with a smaller grid size, and the road condition far from the center of the vehicle is projected into a grid with a larger grid size. While efficiently and accurately detecting negative obstacles, it can effectively reduce the amount of point cloud computing and reduce the complexity of point cloud processing, thereby improving the efficiency of point cloud processing.

[0057] Step 102: Determine the grid label of each grid in the multi-size grid map according to the statistical point cloud distribution characteristics.

[0058] In an embodiment of the present invention, the domain controller traverses the point cloud data in each grid in the multi-size grid map and counts the point cloud distribution characteristics of each grid. The point cloud distribution characteristics include the maximum height, the minimum height and the number of point clouds, and the grid label includes a ground label, a negative obstacle label, a positive obstacle label or an unknown label. The grid label is used to indicate the road condition at the location where the point cloud data corresponding to the grid is located. The ground label is used to indicate that the road condition at the location where the point cloud data corresponding to the grid is located is a relatively flat ground. The negative obstacle label is used to indicate that the road condition at the location where the point cloud data corresponding to the grid is located is a negative obstacle with a certain depth. The positive obstacle label is used to indicate that the road condition at the location where the point cloud data corresponding to the grid is located is a positive obstacle with a certain height. The unknown label is used to indicate that the road condition at the location where the point cloud data corresponding to the grid is located is unknown.

[0059] Step 103: Determine the key feature point pairs of the grid to be identified according to the grid labels and the set feature point pair identification conditions.

[0060] In this step, the grid to be identified includes the grid whose grid label is a negative obstacle label or an unknown label, and the key feature point pair includes the first feature point and the second feature point. The feature point pair identification conditions include that the horizontal distance between points is greater than the horizontal distance between the first adjacent points, and the horizontal distance between points is greater than the horizontal distance between the second adjacent points; the horizontal distance between points is used to indicate the horizontal distance between the first feature point and the second feature point, the horizontal distance between the first adjacent points is used to indicate the horizontal distance between the first feature point and the adjacent points of the first feature point, and the horizontal distance between the second adjacent points is used to indicate the horizontal distance between the second feature point and the adjacent points of the second feature point.

[0061] Optionally, the feature point pair recognition condition further includes a height difference between the first feature point and the second feature point.

[0062] Optionally, the feature point pair recognition condition further includes that a point cloud density at a location where the first feature point is located is different from a point cloud density at a location where the second feature point is located.

[0063] Figure 3 A schematic diagram of a key feature point pair recognition principle provided by an embodiment of the present invention, such as Figure 3As shown in the figure, there are five laser radar points that actually illuminate the ground and the negative obstacle area in the grid to be identified, namely point A, point B, point C, point D, and point E. Among them, point C and point D are in the negative obstacle area, and point C' and point D' are the virtual ground points corresponding to point C and point D respectively. First, the horizontal distance between point B and point C is greater than the horizontal distance between point A and point B, and the horizontal distance between point B and point C is greater than the horizontal distance between point C and point D. Secondly, there is a height difference between point B and point C, indicating that the negative obstacle area has a certain depth. Finally, the point cloud density at the location of point B and point C is different. The point cloud density at the location of point C is greater than the point cloud density at the location of point B, and the point cloud density at the location of point C is greater than the point cloud density at the location of point E, indicating that there are more radar scanning points gathered on the back of the negative obstacle area. Therefore, point B and point C meet the feature point pair recognition conditions, and point B and point C are the key feature point pairs of the grid to be identified.

[0064] Step 104: Perform cluster analysis on the key feature point pairs to identify negative obstacle areas.

[0065] In this step, the domain controller clusters the center points of the key feature point pairs to generate at least one target point cluster; determines whether the number of center points of the target point cluster is greater than a preset number threshold; if it is determined that the number of center points of the target point cluster is greater than or equal to the preset number threshold, the target point cluster is marked as a negative obstacle point cluster; if it is determined that the number of center points of the target point cluster is less than the preset number threshold, the target point cluster is marked as a ground point cluster; and a negative obstacle area is generated based on at least one negative obstacle point cluster. Among them, the center point of the key feature point pair is the midpoint of the first feature point and the second feature point. Specifically, the feature point corresponding to the center point is added to the corresponding set to generate at least one negative obstacle point set, and a negative obstacle area is generated based on at least one negative obstacle point set.

[0066] In an embodiment of the present invention, the target point cluster includes an initial point cluster and a newly created point cluster. The center point of a key feature point pair is randomly selected as the initial node, and the initial point cluster is generated with the initial node as the center; it is determined whether the distance between the center point to be assigned and the initial point cluster is less than a preset distance threshold, and the center point to be assigned is used to indicate the center point outside the initial node; if it is determined that the distance between the center point to be assigned and the initial point cluster is less than the preset distance threshold, the center point to be assigned is added to the initial point cluster; if it is determined that the distance between the center point to be assigned and the initial point cluster is greater than or equal to the preset distance threshold, a new point cluster is generated with the center point to be assigned as the center, until the center points of all key feature point pairs are traversed, so that each center point to be assigned is added to the target point cluster.

[0067] The technical solution provided by the embodiment of the present invention reduces the detection time of negative obstacles on unstructured roads, improves the detection efficiency, and has high real-time performance. At the same time, multiple features are used to detect negative obstacle areas, which reduces the probability of false detection and missed detection, and improves the accuracy and reliability of negative obstacle detection, so that the vehicle can more safely adapt to complex road conditions with negative obstacles, thereby improving the driving safety of the vehicle.

[0068] Figure 4 A flowchart of a grid label determination method provided by an embodiment of the present invention is shown in FIG. Figure 2 As shown, step 102 includes:

[0069] Step 1021, determine whether the number of point clouds in the grid is zero; if it is determined that the number of point clouds in the grid is zero, execute step 1022; if it is determined that the number of point clouds in the grid is not zero, execute step 1023.

[0070] In this step, if it is determined that the number of point clouds in the grid is zero, indicating that there is no point cloud data with reference significance in the grid, and the road condition at the location where the point cloud data corresponding to the grid is located cannot be evaluated, then step 1022 is executed; if it is determined that the number of point clouds in the grid is not zero, indicating that there is point cloud data with reference significance in the grid, then step 1023 is executed.

[0071] Step 1022: Mark the grid label of the grid as an unknown label.

[0072] Step 1023, determine whether the height difference is greater than a preset threshold. If it is determined that the height difference is less than or equal to the preset threshold, execute step 1024; if it is determined that the height difference is greater than the preset threshold, execute step 1025.

[0073] In this step, the height difference is used to indicate the difference between the maximum height and the minimum height.

[0074] Step 1024: Mark the grid labels of the grid as ground labels.

[0075] Step 1025, determine whether the minimum height value is less than zero; if it is determined that the minimum height value is less than zero, execute step 1026; if it is determined that the minimum height value is greater than or equal to zero, execute step 1027.

[0076] Step 1026: Mark the grid label of the grid as a negative obstacle label.

[0077] Step 1027: Mark the grid label of the grid as a positive obstacle label.

[0078] In the embodiment of the present invention, the grids are roughly classified by grid labels, which may cause the grid labels of some grids to be incorrectly marked as negative obstacle labels. However, in the subsequent process, the grids with grid labels as negative obstacle labels will be identified and analyzed again, and the rough classification results will be corrected, thereby further improving the recognition accuracy of negative obstacle areas.

[0079] Figure 5 A schematic diagram of the structure of a negative obstacle area identification device provided by an embodiment of the present invention is shown in FIG. Figure 5 As shown, the device includes a generation module 11, a label determination module 12, a determination module 13 and an identification module 14. The generation module 11 is electrically connected to the label determination module 12, the label determination module 12 is electrically connected to the determination module 13, and the determination module 13 is electrically connected to the identification module 14. The generation module 11 is used to perform rasterization processing on the point cloud data according to the reference distance between the acquired point cloud data and the vehicle to generate a multi-size grid map; the label determination module 12 is used to determine the grid label of each grid in the multi-size grid map according to the statistical point cloud distribution characteristics; the determination module 13 is used to determine the key feature point pairs of the grid to be identified according to the grid labels and the set feature point pair identification conditions; the identification module 14 is used to perform cluster analysis on the key feature point pairs to identify negative obstacle areas.

[0080] In an embodiment of the present invention, the generation module 11 is specifically used to query the grid size corresponding to the reference distance according to the correspondence between the set reference distance and the grid size, divide the point cloud data into grids of the grid size, until all point cloud data are traversed to generate a multi-size grid map.

[0081] In the embodiment of the present invention, the point cloud distribution features include a maximum height, a minimum height, and the number of point clouds.

[0082] In the embodiment of the present invention, the grid label includes a ground label, a negative obstacle label, a positive obstacle label or an unknown label. The label determination module 12 is specifically used to determine whether the number of point clouds in the grid is zero; if it is determined that the number of point clouds in the grid is zero, the grid label of the grid is marked as an unknown label; if it is determined that the number of point clouds in the grid is not zero, it is determined whether the height difference is greater than a preset threshold, and the height difference is used to indicate the difference between the maximum height and the minimum height; if it is determined that the height difference is less than or equal to the preset threshold, the grid label of the grid is marked as a ground label; if it is determined that the height difference is greater than the preset threshold, it is determined whether the minimum height is less than zero; if it is determined that the minimum height is less than zero, the grid label of the grid is marked as a negative obstacle label; if it is determined that the minimum height is greater than or equal to zero, the grid label of the grid is marked as a positive obstacle label.

[0083] In an embodiment of the present invention, the key feature point pair includes a first feature point and a second feature point; the feature point pair recognition condition includes that the horizontal distance between points is greater than the horizontal distance between first adjacent points, and the horizontal distance between points is greater than the horizontal distance between second adjacent points; the horizontal distance between points is used to indicate the horizontal distance between the first feature point and the second feature point, the horizontal distance between first adjacent points is used to indicate the horizontal distance between the first feature point and the adjacent point of the first feature point, and the horizontal distance between the second adjacent points is used to indicate the horizontal distance between the second feature point and the adjacent point of the second feature point.

[0084] In the embodiment of the present invention, the feature point pair recognition condition further includes a height difference between the first feature point and the second feature point.

[0085] In the embodiment of the present invention, the feature point pair recognition condition further includes that the point cloud density at the location where the first feature point is located is different from the point cloud density at the location where the second feature point is located.

[0086] In the embodiment of the present invention, the identification module 14 includes a clustering unit 141, a judgment unit 142, a first marking unit 143, a second marking unit 144 and a generation unit 145. The clustering unit 141 is electrically connected to the judgment unit 142, the judgment unit 142 is electrically connected to the first marking unit 143 and the second marking unit 144 respectively, and the first marking unit 143 is electrically connected to the generation unit 145. The clustering unit 141 is used to cluster the center points of the key feature point pairs to generate at least one target point cluster; the judgment unit 142 is used to judge whether the number of center points of the target point cluster is greater than a preset number threshold; if the judgment unit 142 judges that the number of center points of the target point cluster is greater than or equal to the preset number threshold, the first marking unit 143 is triggered to mark the target point cluster as a negative obstacle point cluster; if the judgment unit 142 judges that the number of center points of the target point cluster is less than the preset number threshold, the second marking unit 144 is triggered to mark the target point cluster as a ground point cluster; the generation unit 145 is used to generate a negative obstacle area according to at least one negative obstacle point cluster.

[0087] In the embodiment of the present invention, the target point cluster includes an initial point cluster and a newly created point cluster. The clustering unit 141 is specifically used to randomly select the center point of a key feature point pair as the initial node, and generate an initial point cluster with the initial node as the center; determine whether the distance between the center point to be assigned and the initial point cluster is less than a preset distance threshold, and the center point to be assigned is used to indicate the center point outside the initial node; if it is determined that the distance between the center point to be assigned and the initial point cluster is less than the preset distance threshold, the center point to be assigned is added to the initial point cluster; if it is determined that the distance between the center point to be assigned and the initial point cluster is greater than or equal to the preset distance threshold, a new point cluster is generated with the center point to be assigned as the center, until the center points of all key feature point pairs are traversed, so that each center point to be assigned is added to the target point cluster.

[0088] The technical solution provided by the embodiment of the present invention reduces the detection time of negative obstacles on unstructured roads, improves the detection efficiency, and has high real-time performance. At the same time, multiple features are used to detect negative obstacle areas, which reduces the probability of false detection and missed detection, and improves the accuracy and reliability of negative obstacle detection, so that the vehicle can more safely adapt to complex road conditions with negative obstacles, thereby improving the driving safety of the vehicle.

[0089] An embodiment of the present invention provides a computer-readable storage medium, which includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute the steps of the embodiment of the above-mentioned negative obstacle area identification method. For a specific description, please refer to the embodiment of the above-mentioned negative obstacle area identification method.

[0090] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A negative obstacle area identification method, characterized in that: The method comprises: rasterizing the point cloud data according to a reference distance between the acquired point cloud data and the vehicle to generate a multi-size raster map; Determine a grid label for each grid in the multi-size grid map according to the statistically determined point cloud distribution characteristics; Determine the key feature point pairs of the grid to be identified according to the grid labels and the set feature point pair identification conditions; Cluster analysis is performed on the key feature point pairs to identify negative obstacle areas.

2. The method according to claim 1, characterized in that The step of rasterizing the point cloud data according to the reference distance between the acquired point cloud data and the vehicle to generate a multi-size grid map includes: According to the correspondence between the set reference distance and the grid size, the grid size corresponding to the reference distance is queried, and the point cloud data is divided into grids of the grid size until all point cloud data are traversed to generate a multi-size grid map.

3. The method according to claim 1, characterized in that The point cloud distribution features include a maximum height, a minimum height, and the number of point clouds.

4. The method according to claim 3, characterized in that The grid labels include ground labels, negative obstacle labels, positive obstacle labels or unknown labels; Determining the grid label of each grid in the multi-size grid map according to the statistical point cloud distribution characteristics includes: Determine whether the number of point clouds in the grid is zero; If it is determined that the number of point clouds in the grid is zero, the grid label of the grid is marked as an unknown label; If it is determined that the number of point clouds in the grid is not zero, then determining whether a height difference is greater than a preset threshold, the height difference being used to indicate a difference between the maximum height and the minimum height; If it is determined that the height difference is less than or equal to the preset threshold, marking the grid label of the grid as a ground label; If it is determined that the height difference is greater than a preset threshold, determining whether the minimum height value is less than zero; If it is determined that the minimum height value is less than zero, the grid label of the grid is marked as a negative obstacle label; If it is determined that the minimum height value is greater than or equal to zero, the grid label of the grid is marked as a positive obstacle label.

5. The method according to claim 1, characterized in that The key feature point pair includes a first feature point and a second feature point; The characteristic point pair recognition condition includes that the horizontal distance between points is greater than the horizontal distance between first adjacent points, and the horizontal distance between points is greater than the horizontal distance between second adjacent points; The horizontal distance between points is used to indicate the horizontal distance between the first feature point and the second feature point, the horizontal distance between the first adjacent points is used to indicate the horizontal distance between the first feature point and the adjacent points of the first feature point, and the horizontal distance between the second adjacent points is used to indicate the horizontal distance between the second feature point and the adjacent points of the second feature point.

6. The method according to claim 5, characterized in that The feature point pair recognition condition further includes a height difference between the first feature point and the second feature point.

7. The method according to claim 6, characterized in that The feature point pair recognition condition also includes that a point cloud density at a location where the first feature point is located is different from a point cloud density at a location where the second feature point is located.

8. The method according to claim 1, characterized in that: The clustering analysis of the key feature point pairs to identify negative obstacle areas includes: Clustering the central points of the key feature point pairs to generate at least one target point cluster; Determine whether the number of center points of the target point cluster is greater than a preset number threshold; If it is determined that the number of center points of the target point cluster is greater than or equal to the preset number threshold, the target point cluster is marked as a negative obstacle point cluster; If it is determined that the number of center points of the target point cluster is less than the preset number threshold, marking the target point cluster as a ground point cluster; A negative obstacle area is generated according to the at least one negative obstacle point cluster.

9. The method according to claim 8, characterized in that The target point cluster includes an initial point cluster and a newly created point cluster; The step of clustering the central points of the key feature point pairs to generate at least one target point cluster includes: Randomly select the center point of a key feature point pair as an initial node, and generate an initial point cluster with the initial node as the center; Determine whether the distance between the center point to be allocated and the initial point cluster is less than a preset distance threshold, the center point to be allocated is used to indicate a center point outside the initial node; If it is determined that the distance between the center point to be allocated and the initial point cluster is less than the preset distance threshold, the center point to be allocated is added to the initial point cluster; If it is determined that the distance between the center point to be assigned and the initial point cluster is greater than or equal to the preset distance threshold, a new point cluster is generated with the center point to be assigned as the center until the center points of all key feature point pairs are traversed, so that each center point to be assigned is added to the target point cluster.

10. A negative obstacle area identification device, characterized in that: The device comprises: A generation module, used for rasterizing the point cloud data according to the reference distance between the acquired point cloud data and the vehicle to generate a multi-size raster map; A label determination module, used to determine the grid label of each grid in the multi-size grid map according to the statistical point cloud distribution characteristics; A determination module, used to determine the key feature point pairs of the grid to be identified according to the grid labels and the set feature point pair identification conditions; The identification module is used to perform cluster analysis on the key feature point pairs to identify negative obstacle areas.

Images