Fast three-dimensional point cloud registration algorithm based on plane and curvature parameters

Abstract The rapid generation of three-dimensional (3D) imaging has improved the safety and operational efficiency of advanced driver assistance systems and mobile robotics technologies. To expedite the stitching process for 3D point cloud data, this study proposes and validates a method that combines coarse registration based on planarity and fine registration based on curvature features. Experimental results demonstrated that compared to the iterative closest point algorithm, the proposed algorithm achieves a faster registration speed, improved by 19.7%, and higher efficiency without compromising the matching residuals. Processing point cloud data without planar information posed challenges in terms of efficiency; however, a notable improvement in processing speed was observed, with point cloud data containing planar scenes. It is important for applications related to 3D imaging registration.