基于微切平面和分块点云密度的点云孔洞边界提取方法OA

Extracting the boundary points of point cloud data is critical for its processing.It improves the quality and efficiency of point cloud processing.This paper proposes a method that combines micro-tangent planes and blocked point cloud density extraction to extract the point cloud boundary.Firstly,this paper constructs the kd-tree spatial topology for point cloud data and selects the sampling points and their k-neighbors as local reference points to align and fit the plane,referred to as the micro-tangent plane,forming a local coordinate system and projecting the k-neighbor points onto the plane.By parameterizing the projected points and summing the field forces of the neighborhood points at the sampling point,preliminary identification of point cloud boundary feature points can be achieved.Secondly,an octree framework is constructed around the point data,facilitating voxelization.Subsequently,the point cloud is partitioned into segments based on the voxel resolution.For each of these segments,the Euclidean distance formula measures the distance from a reference point to its k nearest neighbors.The shortest of these distances is chosen to approximate the segment's average density.A threshold is then established based on the overall point cloud's average density.Any point exceeding this threshold in terms of its own density is excluded,while those falling below are retained.Experimental findings indicate that this algorithm effectively isolates boundary points,even in point clouds with holes.