Statistical analysis of inner products from normal-vectors to 3D point cloud clustering

As a 3D sensor and its application technology had been developed and popularized, technical demands for extraction of significant components from spatial models are increasing. Acquiring spatial information from a point cloud is very important and fundamental technique to process a large number of points. To acquire spatial information from the point cloud, Difference of Normals (DoN) operator and a curvature method have been proposed. However, these methods are influenced by noises or spatial holes. We focused on a slant of a normal vector to obtain a geometric shape from the point cloud. In this paper, we propose statistical analysis to a 3D point cloud with a bivariate frequency distribution of a normal vector inner products. Our experimental results showed that the proposed method is robust to noises or spatial holes. Cluster analysis of this distribution indicated that a bivariate frequency distribution is affected by a local neighborhood.