Research On Point Cloud Processing Technology For Complex Surface Part

With the development of machine vision technology,surface data acquisition of complex surface part can be implemented by three dimensional vision detection based on laser scanning technology,which plays an important role in the digital manufacturing.Since point cloud data collected has characteristics of mass,scatter and redundancy,higher demand is put forward for that CAD model with high precision can be generated by point cloud data efficiently processed.In this paper,some key problems of point cloud processing are studied for complex surface part.First,through the space grid of scattered points cloud data,the topological relationship between point cloud data is reconstructed by searching the k-nearest neighbors of point cloud.On this basis,the normal vector and curvature of the data points are calculated and prepared for the simplification of point cloud data.Secondly,on the basis of analyzing and summarizing the common point cloud simplification algorithm,point cloud data can be simplified in terms of curvature section through improved the simplification algorithm based on curvature,and point cloud simplified can still preserve the corresponding surface features.Finally,aimed at the problem that there are always errors in the point cloud data collected by the laser scanning system,compared with the original CAD model of complex surface part,error compensation algorithm is proposed based on the BP neural network,and the error compensation model is presented.Furtherly,the error compensation experiment is realized by the training samples and test samples of the point cloud data,and the validity of this method can be proved to improve the accuracy of the original point cloud data.On the basis of summarizing the research work in this paper,the prospect of further study is presented on the post-processing of point cloud data for the complex surface part.