Research On Prprocessing Technology Of 3D Laser Poiny Cloud Data In Reverse Engineering

Reverse engineering has developed rapidly in recent years.Its essence is a technology to make product design reappear.Because of its unique advantages,it can greatly shorten the product development cycle,so it has been widely used in many fields.Reverse engineering relies on three-dimensional information acquisition equipment to obtain point cloud data,and then export the collected point cloud data to the experimental equipment to display and preprocess the point cloud data.Finally,three-dimensional reconstruction of the processed point cloud data is carried out.Data preprocessing is the key link of reverse engineering,and the quality of the processing results directly affects the quality of reconstruction.This paper mainly studies the preprocessing of three-dimensional point cloud data,and the related technologies are as follows.Firstly,the research background and application value of reverse engineering and three-dimensional laser point cloud are briefly described.A classification denoising algorithm based on point cloud library is proposed for point cloud denoising.The distance information of point cloud and its neighborhood point cloud is used to remove large-scale noise based on statistical filtering.Then the normal vector information of point cloud and its neighborhood point cloud is used to remove part of small-scale noise based on normal difference filtering.Finally,the residual point cloud is smoothed by bilateral filtering.Compared with other denoising methods,the denoising algorithm in this paper is more sensitive to noise and has better denoising effect.Secondly,an improved point cloud reduction algorithm based on region growing algorithm and curvature estimation is proposed for point cloud reduction.The local distance information,normal information and curvature information of the three-dimensional point cloud model are used to simplify the operation.Firstly,the point cloud data are segmented based on region growing algorithm and curvature estimation.According to the curvature information,the point cloud is divided into different blocks,and then the adaptive simplification operation is carried out for different blocks.The experimental results show that this method retains more feature points without destroying the point cloud model,and it is an ideal algorithm.Finally,an improved ICP algorithm for feature extraction is proposed for point cloud registration.In order to reduce the workload of subsequent registration,the source point cloud is streamlined by the streamlining algorithm in this paper.The normal vector information of point cloud is computed and FPFH feature descriptor is generated.Points with similar feature descriptors in the target point cloud are searched for initial registration based on SAC-IA,and an appropriate initial location is obtained.Finally,ICP registration algorithm based on curvature information feature extraction is implemented.The correctness of this method is verified by experimental analysis.