Research On Point Cloud Registration Optimization Of Complex Surface

With the continuous development of science and technology,many high-precision sensors have come out one after another,and point cloud has gradually become the main data source used to show 3D scenes and parts processing.Influenced by the angle of the sensor and the size of the object,it is very difficult to obtain complete information such as scenes and parts at one time,so the importance of point cloud registration is highlighted.At present,point cloud registration generally adopts the idea of coarse before fine,and this paper also improves the methods of feature point extraction,mismatching point removal and noise point removal based on the idea of coarse before fine,and then uses the improved ICP algorithm for accurate registration,and then optimizes and studies the point cloud registration.The point cloud registration method proposed in this paper improves the registration accuracy of complex surface grinding and polishing precision machining,accelerates the iteration speed,and has good registration effect.At the same time,the method adopted can complete the matching analysis of complex surface parts and theoretical models in the process of complex surface grinding and polishing precision machining,calculate the machining allowance of each position of complex surface grinding and polishing parts,optimize the subsequent machining process parameters and process planning,and further improve the machining accuracy and efficiency of complex surface grinding and polishing.This study is of great significance to the future development of complex surface grinding and polishing.The main work of this paper is as follows:(1)In order to obtain new point cloud data,this paper uses a self-developed three-axis CNC machine tool to process parts,and at the same time uses a Keens contact sensor to scan the surface of parts to obtain point cloud data,which greatly reduces the number of noise points and makes the obtained point cloud data more accurate.(2)Aiming at the problem of noise points in point cloud data,this paper studies statistical filtering,and on this basis,improves it,and puts forward a statistical filtering method based on voxel.The idea of voxel rasterization is introduced into the statistical filtering algorithm.We grid the point cloud data,and then calculate the parameters such as the side length of voxel unit and the number of voxels to improve the statistical filtering algorithm.Then,noise points are introduced and simulated in MATLAB environment,and the removal efficiency of point cloud noise points is improved.(3)For the problem that there are many points in point cloud data,we optimize and improve the ISS(Intrinsic Shape Signature)feature point extraction algorithm,and combine octree with ISS feature point extraction algorithm to extract feature points.The method of extracting feature points by using the Internal Shape Descriptor(ISS)can well preserve the local features around the point cloud data,but the internal points are missing due to the calculation method,and octree as an efficient processing method can extract more accurate feature points,so this paper combines the two methods to extract feature points.Through MATLAB simulation,it is concluded that the optimized algorithm is better than the traditional single algorithm,and the extraction time is also optimized.(4)In order to solve the problem that the rough registration of point clouds can not meet the machining accuracy requirements after the end,this paper deeply studies the ICP algorithm,and based on this,an optimized ICP algorithm is proposed.In this paper,the method of judging the nearest neighbor points is improved,and the method of taking the minimum neighborhood center of gravity formed by three points with minimum Euclidean distance as the nearest neighbor points is put forward.FPFH feature is introduced into the judgment condition,which increases the number of correct matching pairs.MATLAB simulation proves that the accuracy of the improved algorithm is greatly improved and the speed of precise registration is accelerated.