Planar point cloud segmentation based on edge point extraction and redistribution

The main challenge of 3D plane segmentation is to distinguish different planes, and for the density clustering method DBSCAN, the method is invalid if the density is uniform among the planes. To address this shortcoming, this paper proposes a method for edge point extraction and plane matching based on DBSCAN. Firstly, normal vector estimation and connectivity score calculation are performed to distinguish edge points and plane points. DBSCAN clustering is performed on planar points, and then planar matching is performed on edge points based on the clustering results. We conducted experiments on both synthetic and real data. The results show that the method can accurately extract and recognize the edge points with strong robustness, and the method has high accuracy and short computation time.