Research On Key Technology Of 3D Point Cloud Data Processing

Three-dimensional point cloud data processing technology is an important core part of the 3D scanning measurement system,reverse engineering software,computer graphics,computer vision,robot navigation,etc,and has been widely used in many fields.With the development of science and technology,the acquired point cloud data is massive,monolithic and discrete,which must be processed continuously before it can be used in practical application scenarios.The 3D point cloud data processing technology has become the bottleneck to solve the current development problems of related applications,and the research on the theory and method of this technology has been a hot topic for researchers.This paper focused on point cloud data processing technology from point cloud streamlining,point cloud registration and fusion,and point cloud surface reconstruction technology,the specific contents were as follows:1.A point cloud simplification algorithm based on Multi-Angle Threshold preservation boundary points was proposed for the problem of serious loss of boundary points and low precision in traditional point cloud reduction.The boundary points were detected and retained by multi-angle method.The points in non-boundary areas were distinguished by distance Gaussian weighting method,the non-important points were deleted,and the normal vectors were updated to reduce the reduction error.Experiments shown that the algorithm could preserve the boundary feature points and the non-boundary feature points very well.2.A point cloud registration algorithm based on matrix exponent and an improved point cloud fusion algorithm based on moving least squares were proposed for the problems of low accuracy,slow speed in traditional point cloud registration and large error of point cloud fusion.For point cloud registration,the registration objective function was modified to a point-to-plane model,the transformation matrix was expressed by matrix exponential transformation which was solved linearly.The axial encirclement method and the improved moving least square method were used in order to complete point cloud fusion.The experiments shown that the accuracy was high and the speed was fast for point cloud registration,and the smoothness of point cloud fusion was good.3.An implicit surface reconstruction method based on the scale information of point cloud was proposed for the low precision of Poisson surface reconstruction and the problem of generating redundant pseudo-surfaces.The basis function was constructed by the scale distance of each point cloud,and the weight function was constructed by piecewise polynomials.The implicit function was calculated on the adaptive octree voxel,the iso-surface was extracted by the improved algorithm,and then the degraded triangle was removed from the reconstructed surface.Experiments shown that the reconstructed surface had high precision,especially with complex details of the point cloud model.4.The validity and correctness of the proposed algorithms were verified by a comprehensive experiment which carried out with the point cloud data that collected through the laboratory equipment.