A Novel 3D Point Clouds Registration Method Without Data Association

3D point cloud registration is an essential task for localization and mapping technology, in which a correct data association is necessary for point cloud registration algorithms. However, the data association is time-consuming and prone to mismatching, which degrades the efficiency and accuracy of point cloud registration. In this paper, to solve this problem, a direct 3D point cloud registration method without data association is proposed, in which the relative pose is achieved through extracting and aligning the principal directions of two point-cloud sets. The point clouds are projected into the three principal directions and the variances of projected points are compared to obtain the unique solution. The proposed algorithm can quickly and accurately register any two sets of point cloud data for the same object collected at different locations. Simulation results show that the proposed method can achieve 3D point cloud registration more efficiently and accurately than traditional methods based on data association.