Research On 3D Point Cloud Segmentation Based On Geometric Features

Point cloud segmentation is a hot and difficult problem in 3D point cloud processing,and is also challenging research topic.Existing edge-based point cloud segmentation methods are prone to over-segmentation or under-segmentation problems,and segmentation boundaries are prone to aliasing.Therefore,this paper proposes a three-dimensional scattered point cloud segmentation method based on geometric features,which can better solve the problem of over-segmentation and under-segmentation,optimize the segmentation boundary,and improve the segmentation accuracy of point cloud.The main work of this paper is as follows:(1)For the complex point cloud model,the edge features are not obvious,and the edge feature description sub-description ability is insufficient.We propose a feature description method of local weighted curvature to enhance the expression ability of curvature features.Based on the feature description,the sharp feature points can be effectively extracted,which improves the accuracy of edge feature point extraction.In addition,this paper improves the neighbor search algorithm in the process of local space construction,solving the problem of low search efficiency and inaccurate feature description caused by insufficient distance constraint.The experimental results show that the proposed method can extract edge feature points efficiently and accurately,and has good robustness.(2)For the point cloud model extraction of concave edge features in the process of single curvature information or the extraction of concave edge features based on over-segmentation results,the concave edge feature extraction is not accurate enough.This paper proposes a concavity feature extraction algorithm based on grid point cloud data.According to the angle between adjacent grids and the adjacent two grids,the volume characteristics of the tetrahedron are formed,and the geometrical characteristics between the normal vector and the distance vector between two adjacent grids are constrained to effectively extract the concave edge feature points.The experimental results show that the proposed algorithm can accurately extract concave edge feature points representing 3D point clouds.(3)For the problems of over-segmentation and edge-saw,this paper proposes a segmentation algorithm based on geometric feature local consistency constraint.Firstly,the point cloud is initially segmented by direction weighting and distance constraint mechanism to obtain the initial segmentation edge.Based on the local boundary iterative optimization method proposed,the local neighborhood of the initial segmentation edge is subdivided,and the segmentation accuracy is improved.The evaluation results based on the standard evaluation indicators show that the proposed algorithm is effective.