CN-122023431-A - Ground segmentation method, equipment, medium and product based on roadside laser radar

CN122023431ACN 122023431 ACN122023431 ACN 122023431ACN-122023431-A

Abstract

The embodiment of the application relates to the field of intelligent traffic and discloses a ground segmentation method, equipment, medium and product based on a road side laser radar; the method comprises the steps of obtaining point cloud data collected by a laser radar, carrying out rasterization processing on the point cloud data, selecting the lowest point in each grid as an initial seed point, calculating the statistical characteristic values of the heights of all the initial seed points, determining effective seed points based on the statistical characteristic values, determining the weight value of each effective seed point according to the distance between the effective seed points and a road side laser radar, carrying out plane fitting processing based on the effective seed points and the weight values corresponding to the effective seed points, and constructing a ground model to complete ground segmentation, so that the accuracy of a ground segmentation algorithm under long distance and large scenes is remarkably improved.

Inventors

MA YUAN
XUAN ZHIYUAN
YANG XUAN
ZHANG LIJUAN

Assignees

云控智行科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260119

Claims (10)

1. A ground segmentation method based on a roadside laser radar, the method comprising: Acquiring point cloud data acquired by a laser radar, performing rasterization on the point cloud data, and selecting the lowest point in each grid as an initial seed point; Calculating the statistical characteristic values of the heights of all the initial seed points, and determining effective seed points based on the statistical characteristic values; determining a weight value of each effective seed point according to the distance between the effective seed point and the roadside laser radar; And executing plane fitting processing based on the effective seed points and the corresponding weight values thereof, constructing a ground model and completing ground segmentation.
2. The method of claim 1, wherein the step of acquiring point cloud data acquired by the lidar, rasterizing the point cloud data, and selecting a lowest point in each grid as an initial seed point specifically comprises: counting coordinate extremum of the point cloud data in the horizontal dimension, and calculating the width and the height of the grid map by combining with the preset grid size; And constructing an initialized grid coordinate matrix based on the width and the height, wherein the grid coordinate matrix is used for recording the height of the lowest point corresponding to each grid position and the corresponding original point index.
3. The method of claim 2, wherein the step of acquiring point cloud data acquired by the lidar, rasterizing the point cloud data, and selecting a lowest point in each grid as an initial seed point further comprises: traversing the point cloud data, and calculating a grid index corresponding to the current data point in the grid coordinate matrix according to the coordinate extremum and the grid size; Acquiring the height of the current data point, and comparing the height with the lowest point height stored in the grid position corresponding to the grid index; If the height of the current data point is smaller than the height of the lowest point, updating the height of the lowest point of the grid position to be the height of the current data point, and updating the corresponding original point index; and after traversing, extracting points corresponding to all original point indexes recorded in the grid coordinate matrix to serve as the initial seed points.
4. The method according to claim 1, wherein the step of calculating statistical eigenvalues of the heights of all the initial seed points and determining valid seed points based on the statistical eigenvalues comprises in particular: Calculating the mean mu and variance sigma of all initial seed point heights; using the formula Calculating the height threshold , wherein, Is a preset coefficient; will be less than the height threshold And (3) reserving the rest points as the effective seed points.
5. The method of claim 1, wherein determining the weight value for each of the valid seed points based on the distance of the valid seed points relative to the roadside lidar comprises calculating the weight value using the formula : Wherein, the method comprises the following steps of , , ) Is the coordinates of the i-th valid seed point, The furthest distance of interest for a preset seed point.
6. The method according to claim 1, wherein the step of performing a plane fitting process based on the valid seed points and their corresponding weight values to construct a ground model to complete ground segmentation specifically comprises: randomly selecting a set number of sample points from the effective seed points to estimate a current ground model; Calculating the geometric distance from the effective seed point to the current ground model; And calculating a cost value by combining the geometric distance and the weight value, and judging the effective seed point as an inner point if the cost value is smaller than a preset threshold value.
7. The method of claim 6, wherein the method further comprises: iteratively executing the selected set number of sample points to estimate a current ground model, calculate a cost value and judge an interior point; After each iteration, counting the number of interior points corresponding to the current ground model, if the number of interior points is larger than the maximum number of interior points recorded currently, updating the maximum number of interior points, and temporarily storing the current ground model as a candidate ground model; judging whether a preset iteration termination condition is met, if so, stopping iteration, and outputting the candidate ground model as a final ground model, wherein the iteration termination condition comprises that the proportion of the number of the inner points to the total number of the effective seed points reaches a proportion threshold value or the iteration times reaches a preset frequency threshold value.
8. An electronic device, the electronic device comprising: One or more processors, and A memory storing computer program instructions that, when executed, cause the processor to perform the steps of the method of any one of claims 1 to 7.
9. A computer readable medium having stored thereon a computer program/instruction, which when executed by a processor, implements the steps of the method according to any of claims 1 to 7.
10. A computer program product comprising computer programs/instructions which, when executed by a processor, implement the steps of the method of any of claims 1 to 7.

Description

Ground segmentation method, equipment, medium and product based on roadside laser radar Technical Field The application relates to the field of intelligent traffic, in particular to a road side laser radar-based ground segmentation method, equipment, medium and product. Background Along with the development of intelligent traffic and automatic driving technologies, the roadside laser radar is widely applied to traffic monitoring and vehicle-road cooperative systems as important sensing equipment. In the point cloud data collected by the roadside lidar, the ground point cloud generally occupies a considerable proportion, typically between 20% and 40%. The large number of ground points not only increases the calculation burden of the subsequent target detection algorithm, which results in increased time consumption, but also is easy to be erroneously detected as an obstacle target in a complex scene, thereby affecting the decision security of the unmanned vehicle. In addition, when multi-frame point cloud matching or feature extraction is performed, the reserved ground points can also generate certain interference on the matching precision. Therefore, accurately and efficiently dividing and removing the ground point cloud in the preprocessing stage is an indispensable key step in the laser radar data processing flow. In the prior art, a plane fitting method based on random sample consensus (RANSAC) is often used to remove ground points. The method generally selects a plurality of points with the lowest height as seed points, and then fits a ground plane model. However, in conventional seed point selection strategies, a "one-shot" approach is often simply employed, i.e., all points below a certain fixed height threshold are selected as seed points. The method has obvious limitations in road side perception scenes, namely, on one hand, the installation of the road side laser radar often has a certain inclination angle or the ground has a gradient, so that a fixed height threshold value is difficult to adapt to the whole scene, and on the other hand, reflection noise points (such as underground false points caused by water pit reflection or system errors) below the ground are often existed in actually collected data, and can be mistakenly selected as seed points, so that the accuracy of plane fitting is seriously affected. In addition, as the distance increases, the point cloud becomes sparse, the bottom of distant objects (such as the vehicle chassis or the wall root) may be misjudged as the lowest point of the area, and if these unreliable distant points and near points have the same fitting weights, the fitted ground model will deviate from the real ground. Disclosure of Invention The application aims to provide a ground segmentation method, equipment, medium and product based on a road side laser radar, which are at least used for solving the problem that a ground model fitting failure is caused by the fact that a seed point selection method in the prior art depends on a fixed height threshold value and is easy to mistakenly select a noise point as a seed point. To achieve the above object, some embodiments of the present application provide the following aspects: In a first aspect, some embodiments of the present application provide a ground segmentation method based on a roadside lidar, the method comprising: Acquiring point cloud data acquired by a laser radar, performing rasterization on the point cloud data, and selecting the lowest point in each grid as an initial seed point; Calculating the statistical characteristic values of the heights of all the initial seed points, and determining effective seed points based on the statistical characteristic values; determining a weight value of each effective seed point according to the distance between the effective seed point and the roadside laser radar; And executing plane fitting processing based on the effective seed points and the corresponding weight values thereof, constructing a ground model and completing ground segmentation. In a second aspect, some embodiments of the application also provide an electronic device comprising one or more processors and a memory storing computer program instructions that, when executed, cause the processors to perform the steps of the method as described above. In a third aspect, some embodiments of the application also provide a computer readable medium having stored thereon computer program instructions executable by a processor to implement a method as described above. In a fourth aspect, some embodiments of the application also provide a computer program product comprising a computer program/instruction which, when executed by a processor, implements the steps of the method as described above. Compared with the related art, in the scheme provided by the embodiment of the application, firstly, the dynamic threshold value is determined by counting the statistical characteristic values of the heights of all initial see