A Point Cloud Registration Algorithm Based on Planar Projection Contour

This paper proposes a novel registration algorithm based on the planar projection contour descriptor (PPC). The construction process of the PPC is as follows. Firstly, a local coordinate system is established based on the query point and its neighboring points' information. Secondly, the three-dimensional information of the neighboring points is projected onto a three-coordinate plane to obtain two-dimensional information. Then, the projection point set is fitted into a convex hull contour, which is encoded as the local feature of the plane. Finally, the local features of the three planes are concatenated in an orderly fashion to form PPC. In addition, the PPC-based point cloud registration algorithm was tested and compared with classical algorithms using the Stanford dataset. The results showed that PPC can accurately identify the feature matching relationship between the source and target point clouds, leading to precise registration. This demonstrates the superiority of the algorithm.