关键词: 稀疏点云, 点云配准, 四点算法, ICP算法

Abstract: The point cloud data registration is one of the key technologies of threedimensional reconstruction. To solve the registration issue of sparse point cloud scanned from the noncooperative spacecraft, we propose a improved 4points congruent sets (4PCS) algorithm to obtain the preliminary registration result, and optimize the final alignment with the improved iterated closest points (ICP) algorithm. Then，a novel point cloud simplification algorithm using uniform sampling is proposed based on KDTree. The uniform relation of the threshold parameters is established via the density of the point cloud. The results show that the proposed algorithm can effectively achieve good alignments of the sparse point cloud of the satellite, and the consuming time is decreased to 42.49% compared with the Geometric Hashing algorithm.

Key words: noncooperative target, sparse point cloud, point cloud registration, 4PCS algorithm, ICP algorithm