Point Cloud Registration Algorithm Based on Bounded Rotation Angle

In the paper, a registration algorithm based on bounded rotation angle is proposed to solve the problem of the existing 3D point cloud algorithm which can't complete registration with large rotation angle change. Firstly, the global registration is transformed into regional registration according to the low dimensional nature of the rigid body transformation, and it achieves initial registration through the four steps of region division, region registration, solving combination coefficient and solving rigid body transformation; Secondly, the basic ICP algorithm is improved by setting boundary for rotation angle to achieve fine registration. The experimental results show that the proposed algorithm in the paper is accurate, fast and robust, and it can avoid the interference of noises and outliers.