Research on lidar target detection for metro scene

Aiming at the problems of difficult real-time construction of train running forward boundary and complex cluster detection of unstructured track and road environment targets, a lidar based on target detection and forward boundary real-time construction system was proposed. Firstly, the strategy of space for time was adopted to improve the efficiency of rail track extraction, and the rail track was extracted steadily by using the local elevation information, global geometric parallel information and multi-frame information of the rail track. Then, the forward limit of train passage was constructed based on the extracted rail track position in real time. Secondly, based on the point cloud depth map and gradient and distance characteristics, the obstacle clustering problem in non-paved road scenarios was solved, and the track information was used for multi-target tracking to improve the stability of target tracking. By real train test on metro scene, it was proved that the proposed method could extract track position accurately and reliably, it also could detect and track targets.