Research on Lidar Point Cloud Segmentation and Collision Detection Algorithm

Abstract

3D laser radar is widely used in unmanned driving systems due to its high precision, strong anti-interference ability and omni-directional scanning. It is used for road environment detection and anti-collision detection. In order to increase the accuracy of the laser radar to segment the scanning point cloud, firstly, voxel filtering is applied to the point cloud to reduce the number of point clouds, and then the point cloud is removed by using the point cloud progressive morphology filtering method. Finally, the point cloud is divided into several independent clusters by the region-growing algorithm, and the segmented point cloud clusters are generated into a bounding box. By detecting the relative positional relationship of the bounding box of the three-dimensional space, it is determined whether there is a collision of the two physical models. The experimental results show that after the point cloud preprocessing and clustering segmentation, the collision detection of the model combined with the bounding box can effectively identify and accurately determine the spatial positional relationship of the object and improve the accuracy of collision recognition.