A Unified Detection Method of Boundary and Hole Points Based on Point Cloud Resolution

Abstract

Target self-occlusion and view limitation of scanning equipment will make boundary and hole points exist in 3D point cloud data. The detection of boundary and hole points is an essential processing step for many related applications, such as the repair of point cloud holes. In this paper, we propose a unified detection algorithm for the point cloud's boundary and hole points. First, the algorithm utilizes the point cloud resolution to define a spherical neighbourhood and establishes the covariance matrix through neighbours. Then, we analyze the covariance matrix and invoke the angle criterion to judge whether the data point is a boundary point. The experimental results are carried out, which indicates that the proposed approach is superior to the comparative methods in time consumption and the success rate of boundary point detection. Furthermore, it is also robust to changes in geometric structure and the number of holes in the point cloud.