激光雷达点云配准技术研究

LiDAR is an important sensor in the fields of three-dimensional positioning, navigation, mapping and autonomous driving. Point cloud registration technology is an indispensable and important pre-step for processingpoint cloud data of LiDAR. This article studies the commonly used methods in the field of LiDARpoint cloud registration, and summarizes the connections between them according to different classification standards. Subsequently, according to the different processing primitives, a lot of methods are divided into point primitives, geometric primitives and voxel primitives based method families. The basic theoryand standard algorithm procedures of different familiesof methods are refined, part of the important modules are introduced and analyzed, and the advantages and disadvantages of various methods are analyzed and compared. Finally, an important issue in the explorationof the point cloud registration algorithm, the registration evaluation system is analyzed in depth, and the possible future improvement directions of the point cloud registration method are prospected.