A method for detecting the road surface and retaining walls on both sides of an unmanned mining truck driving scene

Abstract

The present invention discloses a method for detecting the road surface and retaining walls on both sides of an unmanned mining truck driving scene. The method comprises obtaining a triangular mesh model of the unmanned mining truck driving scene; performing plane fitting on the point cloud data within the triangular mesh model; calculating the distance from the point cloud data to the plane within the plane, and calculating the variance of the plane distance; comparing the variance with a variance threshold, and dividing the triangular mesh model into a junction grid and a target grid based on the comparison result; distinguishing road surface points and retaining wall points within the junction grid by calculating the mutation points within the junction grid; and distinguishing road surface points and retaining wall points within the target grid by calculating the slope value of the target grid. The method proposed by the present invention can effectively detect the road surface and retaining walls on both sides in mining scenarios with complex road conditions.

Description

Technical Field

[0001] The present invention relates to the field of unmanned driving, and in particular to a method for detecting a road surface and retaining walls on both sides of an unmanned mining truck driving scene. Background Art

[0002] With the continuous development of autonomous driving technology, it has gradually been applied to mining areas. Throughout the autonomous driving system, the perception module serves as the system's "eyes." Accurately detecting the road surface and retaining walls within the driving scene of the unmanned mining truck is a crucial perception preprocessing step. Because LiDAR point cloud data is unevenly distributed and discrete, road surface detection often suffers from over-segmentation and under-segmentation. Inaccurate detection results can lead to false detection of obstacles and misdetection of accessible areas, posing a significant challenge for autonomous driving systems. Good road surface and retaining wall detection results can reduce subsequent processing time and improve overall system performance. Therefore, accurate road surface and retaining wall detection is crucial. Summary of the Invention

[0003] The purpose of the present invention is to provide a method for detecting roads and retaining walls on both sides in a mining area scene, so as to solve the problem of unmanned mining trucks performing road and retaining wall detection in a mining area scene.

[0004] The present invention provides a method for detecting roads and retaining walls on both sides in a mining area, comprising:

[0005] Obtain a triangular mesh model of the unmanned mining truck driving scene;

[0006] Performing plane fitting on the point cloud data in the triangular mesh model;

[0007] Calculate the distance from the point cloud data in the plane to the plane, and calculate the variance of the plane distance;

[0008] Comparing the variance with a set variance threshold, and dividing the triangular mesh model into a boundary mesh and a target mesh according to the comparison result;

[0009] By calculating the mutation points in the boundary grid, the road surface points and retaining wall points in the boundary grid are determined;

[0010] The road surface points and retaining wall points within the target grid are identified by calculating the slope value of the target grid.

[0011] Furthermore, the method for constructing the triangular mesh model includes:

[0012] Obtaining point cloud data in the unmanned mining truck driving scene and filtering out noise from the point cloud data;

[0013] Rasterizing the point cloud data and extracting the highest point in the grid as a feature point of the unmanned mining truck driving scene;

[0014] Construct a triangular mesh model based on the feature points of the unmanned mining truck driving scene.

[0015] Furthermore, the method for filtering noise from the point cloud data includes:

[0016] According to the point cloud coordinates in the unmanned driving scene, determine whether the points in the point cloud data are invalid points. If so, treat the invalid points as noise points and filter them out.

[0017] Alternatively, the number of adjacent points of any point in the point cloud of the unmanned driving scene within a radius of r is calculated. If the number is less than a set number threshold, the point is regarded as a noise point;

[0018] Alternatively, the average distance between any point in the point cloud of the unmanned driving scene and its k adjacent points is calculated. If the average distance is less than a set distance average threshold, the point is treated as a noise point and filtered out.

[0019] Furthermore, the method of rasterizing the point cloud data and extracting the highest point in the grid as a feature point of the unmanned mining truck driving scene includes:

[0020] Based on the right-hand rule, the front of the vehicle is the x-axis, the left direction of the vehicle is the y-axis, and the top of the vehicle is the z-axis;

[0021] The point cloud data of the unmanned mining truck driving scene acquired by the LiDAR is divided into several small grids along the x-axis and y-axis in the xoy plane;

[0022] The point with the highest z value projected in each grid is used as the feature point of the unmanned mining truck driving scene.

[0023] Furthermore, the method of constructing a triangular mesh model based on the feature points of the unmanned mining truck driving scene includes:

[0024] Determine the positions of the four corner points in the xoy plane in the point cloud data of the unmanned mining truck driving scene, and select any three corner points to generate a triangle as the initial triangular mesh model;

[0025] The extracted feature points of the unmanned mining truck driving scene are traversed and added to the triangular mesh model of the unmanned mining truck driving scene, and a new triangular mesh model is generated by newly creating and reorganizing triangles.

[0026] Furthermore, the method for plane fitting of point cloud data in the triangular mesh includes:

[0027] The point cloud data of the unmanned mining truck driving scene is traversed and projected into a triangular mesh model. The RANSAC method is used to obtain the fitting plane of the point cloud data in the triangular mesh model.

[0028] Furthermore, the method of comparing the variance with the variance threshold and dividing the triangular mesh model into the boundary mesh and the target mesh includes:

[0029] Compare the obtained variance with the set variance threshold;

[0030] According to the set variance threshold, the triangular mesh is divided into a boundary mesh with a variance greater than the variance threshold and a target mesh with a variance less than the variance threshold.

[0031] Furthermore, the method for calculating the mutation point in the boundary grid includes:

[0032] The point cloud data of the unmanned mining truck driving scene within the intersection grid are sorted according to the z value, and traversed from the lowest point in sequence. The z value height difference between the current point and the next adjacent point is calculated. If it is greater than the set height difference threshold, the current point is regarded as a mutation point.

[0033] Furthermore, the method for distinguishing road surface points and retaining wall points within the boundary grid includes:

[0034] Traverse the z value of each point in the intersection grid. If the z value is less than the z value of the mutation point, the point is considered a road surface point, otherwise it is a retaining wall point.

[0035] Furthermore, the formula for calculating the slope value of the target grid is as follows:

[0036] ;

[0037] Where: Vec_x, Vec_y, and Vec_z represent the x, y, and z coordinate values ​​of the normal vector coordinates calculated from the point cloud fitting plane within the target grid.

[0038] Furthermore, the method for distinguishing road surface points and retaining wall points within the target grid includes:

[0039] The slope values ​​in the target grid are traversed. If the slope value is less than the slope threshold, the point in the target grid is considered a road surface point, otherwise it is a retaining wall point.

[0040] The beneficial effects of the present invention are as follows:

[0041] The method proposed in this invention can effectively detect roads and retaining walls on both sides in mining scenarios with complex road conditions.

[0042] The method proposed in this invention uses the highest point in the grid as the feature point to construct a triangular mesh model. Unlike airborne radar, the data of unmanned mining trucks in mining areas are not divided into such large scales and there is no large amount of vegetation as an obstruction. Therefore, choosing the highest point as the feature point is more in line with the actual terrain characteristics.

[0043] The method proposed in this invention uses lidar data to directly construct a triangular mesh model. It does not rely on special radar equipment or previous DEM model parameters. It can effectively build a network based on the actual terrain and is suitable for complex terrain scenes.

[0044] The method proposed in the present invention uses slope features and mutation points to detect roads and retaining walls, which is more robust than using absolute height differences for discrimination. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A schematic flow chart of a method for detecting a road surface and retaining walls on both sides of an unmanned mining truck driving scenario according to an embodiment of the present invention;

[0046] Figure 2 A schematic diagram of triangular mesh construction in a method for detecting a road surface and retaining walls on both sides of an unmanned mining truck driving scenario provided by an embodiment of the present invention;

[0047] Figure 3 A schematic diagram of calculating mutation points after statistical sorting of z-values ​​of point clouds within a boundary grid in a method for detecting a road surface and retaining walls on both sides of an unmanned mining truck driving scene provided by an embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram of the calculation of the slope value by calculating the normal vector of the point cloud in the target grid through plane fitting in a method for detecting the road surface and retaining walls on both sides of an unmanned mining truck driving scene provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The present invention proposes a method for detecting the road surface and the retaining walls on both sides of the unmanned mining truck driving scene. Figure 1 As shown, the method includes the following steps:

[0050] S1: For each point in the point cloud data of the unmanned mining truck driving scene acquired by the lidar, calculate the average distance from it to all adjacent k points and the number of adjacent points within a radius of r. If there is a NaN value (invalid value) in the three-dimensional coordinates of the point or the above average distance is less than the set distance average threshold or the number of adjacent points is less than the set number threshold, the point is considered to be a noise point and is filtered out. This application does not limit the method for filtering out noise points, and any filtering scheme is within the scope of protection of this application.

[0051] S2: In the xoy plane, along the x-axis and y-axis directions, the point cloud data of the unmanned mining truck driving scene is rasterized. The grid length is set according to the actual site size. The 3D coordinates of the point cloud are used to calculate the grid to which it belongs. At the same time, the point with the largest z value in each grid is found as the feature point of the unmanned mining truck driving scene.

[0052] S3: First, determine the four corner points of the point cloud data of the unmanned mining truck driving scene in the xoy plane, and select any three corner points to generate a triangle. Figure 2 As shown in (a), the solid line represents the first triangle we constructed; then, using the feature points of the unmanned mining truck driving scene obtained in step S2, we sequentially add them to the triangular mesh model of the unmanned mining truck driving scene, traverse the existing triangles, and determine whether the insertion point is located inside a triangle. If the insertion point is inside the triangle, we generate three new edges from the vertex to the insertion point, as shown in Figure 2 (b) If the insertion point is outside the triangle, directly add two edges to generate a new triangle, as shown in Figure 2 (c) shown.

[0053] S4: Perform RANSAC plane fitting on the point cloud data of the unmanned mining truck driving scene in each triangular mesh, obtain the fitting plane equation and calculate the normal vector Vec (Vec_x, Vec_y, Vec_z) of the point cloud in the triangular mesh;

[0054] S5: Calculate the distance value from the point cloud to the plane in each triangular plane, and calculate the variance of the distance value, which represents the distribution characteristics of the point cloud.

[0055] S6: Compare the variance obtained in step S5 with the set threshold. If the variance is greater than the set variance threshold, it means that the fitting plane is significantly different from the actual terrain and may be in two or more planes. The point is considered to be at the junction of the road surface and the retaining wall. If the variance is less than the set variance threshold, the point is considered to be in the same plane. According to the set variance threshold, the triangular mesh is divided into a junction mesh with a variance greater than the set variance threshold and a target mesh with a variance less than the set variance threshold.

[0056] S7: As Figure 3 As shown in the figure, the point cloud data of the unmanned mining truck driving scene within the intersection grid is sorted according to the z value, and traversed from the lowest point in sequence. The z value height difference between the current point and the next point is calculated. If it is greater than the set height difference threshold, the current point is regarded as a mutation point, that is, the current point is considered to be the intersection point of the road surface and the retaining wall.

[0057] S8: Figure 4 As shown, the slope value of the grid plane is calculated by fitting the normal vector of the plane, as shown in the formula:

[0058] .

[0059] S9: In the boundary grid, if the z value of a point is less than the z value of the mutation point, the point is considered to be a road surface point, otherwise it is a retaining wall point.

[0060] S10: In the target grid, the slope value of the grid plane is compared with the set slope threshold. Points in the target grid that are less than the threshold are considered to be road surface points, otherwise they are considered to be retaining wall points.

[0061] The present invention uses data from mining areas to conduct experiments, which can better perform road surface detection.

[0062] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for detecting the road surface and retaining walls on both sides of an unmanned mining truck driving scene, characterized in that: include: Obtain a triangular mesh model of the unmanned mining truck driving scene; Performing plane fitting on the point cloud data in the triangular mesh model; Calculate the distance from the point cloud data in the plane to the plane, and calculate the variance of the plane distance; Comparing the variance with a set variance threshold, and dividing the triangular mesh into a boundary mesh having a variance greater than the set variance threshold and a target mesh having a variance less than the set variance threshold according to the comparison result; By calculating the mutation points in the boundary grid, the road surface points and retaining wall points in the boundary grid are determined; The road surface points and retaining wall points within the target grid are identified by calculating the slope value of the target grid.

2. The method for detecting the road surface and retaining walls on both sides of the unmanned mining truck driving scene according to claim 1 is characterized in that: The method for constructing the triangular mesh model includes: Obtaining point cloud data in the unmanned mining truck driving scene and filtering out noise from the point cloud data; Rasterizing the point cloud data and extracting the highest point in the grid as a feature point of the unmanned mining truck driving scene; Construct a triangular mesh model based on the feature points of the unmanned mining truck driving scene.

3. The method for detecting the road surface and retaining walls on both sides of the unmanned mining truck driving scene according to claim 2 is characterized in that: The method for filtering noise from the point cloud data includes: According to the point cloud coordinates in the unmanned driving scene, determine whether the points in the point cloud data are invalid points. If so, treat the invalid points as noise points and filter them out. Alternatively, the number of adjacent points of any point in the point cloud of the unmanned driving scene within a radius of r is calculated. If the number is less than a set number threshold, the point is regarded as a noise point; Alternatively, the average distance between any point in the point cloud of the unmanned driving scene and its k adjacent points is calculated. If the average distance is less than a set distance average threshold, the point is treated as a noise point and filtered out.

4. The method for detecting the road surface and retaining walls on both sides of the unmanned mining truck driving scene according to claim 2 is characterized in that: The method of rasterizing the point cloud data and extracting the highest point in the grid as a feature point of the unmanned mining truck driving scene includes: Based on the right-hand rule, the front of the vehicle is the x-axis, the left of the vehicle is the y-axis, and the top of the vehicle is the z-axis; The point cloud data of the unmanned mining truck driving scene acquired by the LiDAR is divided into several small grids along the x-axis and y-axis in the xoy plane; The point with the highest z value projected in each grid is used as the feature point of the unmanned mining truck driving scene.

5. The method for detecting the road surface and retaining walls on both sides of the unmanned mining truck driving scene according to claim 2 is characterized in that: The method of constructing a triangular mesh model based on the feature points of the unmanned mining truck driving scene includes: Determine the positions of the four corner points in the xoy plane in the point cloud data of the unmanned mining truck driving scene, and select any three corner points to generate a triangle as the initial triangular mesh model; The extracted feature points of the unmanned mining truck driving scene are traversed and added to the triangular mesh model of the unmanned mining truck driving scene, and a new triangular mesh model is generated by newly creating and reorganizing triangles.

6. The method for detecting the road surface and retaining walls on both sides of the unmanned mining truck driving scene according to claim 1 is characterized in that: Methods for plane fitting of point cloud data within a triangular mesh include: The point cloud data of the unmanned mining truck driving scene is traversed and projected into a triangular mesh model. The RANSAC method is used to obtain the fitting plane of the point cloud data in the triangular mesh model.

7. The method for detecting the road surface and retaining walls on both sides of the unmanned mining truck driving scene according to claim 1 is characterized in that: Methods for calculating the discontinuity points within the boundary grid include: The point cloud data of the unmanned mining truck driving scene within the intersection grid are sorted according to the z value, and traversed from the lowest point in sequence. The z value height difference between the current point and the next adjacent point is calculated. If it is greater than the set height difference threshold, the current point is regarded as a mutation point.

8. The method for detecting the road surface and retaining walls on both sides of the unmanned mining truck driving scene according to claim 1 is characterized in that: Methods for distinguishing road surface points and retaining wall points within the boundary grid include: Traverse the z value of each point in the intersection grid. If the z value is less than the z value of the mutation point, the point is considered a road surface point, otherwise it is a retaining wall point.

9. The method for detecting the road surface and retaining walls on both sides of the unmanned mining truck driving scene according to claim 1 is characterized in that: The formula for calculating the slope value of the target grid is as follows: ； Where: Vec_x, Vec_y, and Vec_z represent the x, y, and z coordinate values ​​of the normal vector coordinates calculated from the point cloud fitting plane within the target grid.

10. The method for detecting the road surface and retaining walls on both sides of the unmanned mining truck driving scene according to claim 1 is characterized in that: Methods for distinguishing road surface points and retaining wall points within the target grid include: The slope values ​​in the target grid are traversed. If the slope value is less than the slope threshold, the point in the target grid is considered a road surface point, otherwise it is a retaining wall point.

Images