Moving target removal for lidar SLAM

Lidar SLAM is mainly suitable for static scenes without moving targets. The existence of motion in actual scenes will limit the application of the algorithm. If there are moving targets in the lidar field of view, the 3D point cloud map constructed by the lidar SLAM will include a series of fake point cloud from the moving target. This situation leads to insufficient accuracy of the constructed map and affects the applications such as navigation and route planning. In this paper, a moving target removal algorithm is proposed, which realizes the detection and removal of moving targets through multi-frame accumulation, change detection, and point cloud clustering of point clouds. The experiments base d on measured lidar point cloud show that the proposed algorithm can optimize the mapping results of the existing lidar SLAM, remove the continuous trajectory caused by the moving target, and construct a more accurate 3D map.