Obstacle Detection and Tracking of Unmanned Mine Car Based on 3D Lidar

The current open-pit mine environment is complex, and mine car transportation accidents occur frequently. In order to solve the safety problems in production, this paper proposes an obstacle detection and tracking method based on 3D lidar. After the original point cloud data is obtained, the point cloud data is sliced, and the real-time map is obtained through the fixed route path, and the region of interest is extracted. we segment the ground point cloud within the data of the region of interest and clustering the non-ground point cloud using distance-based obstacle clustering algorithm to obtain the location and size of the obstacle. Since there are many obstacles such as people, cars, and soil slopes in the mining area, and most of the weather is dusty, which will increase the false discovery rate of the detection results. the obstacle tracking algorithm based on improved Hungary matching is used to perform similarity between the obstacles and the tracks in the front and rear frames, output the obstacles that are continuously matched successfully, and stabilize the output conditions of the obstacles. An unmanned mining vehicle experimental platform was built in the experimental field, and unmanned vehicle experiments under different scenarios were carried out. Experimental results show that the perception algorithm can accurately detect obstacles of 70*70cm, 50*100cm, and 50*150cm on the road. Obstacle tracking filtering based on the improved Hungarian matching algorithm can detect obstacles on the road stably. The frame loss rate has been reduced from 20% to 10%, and the false discovery rate of obstacles has been reduced from the original 32% to 14%, the stability and reliability have been greatly improved, which initially meets the actual application requirements.