The Research On Feature Extraction Of Edge And Surface From 3D Point Cloud Of Industrial Components

With the rapid development of computer-aided design,reverse engineering and other computer technologies,the design and development of industrial components incorporate a lot of new digital technologies.The model of digitized industrial components makes it available to digital measurement and shape analysis,which can quickly realize the process of being from entity to design drawings as well.Besides that,it can also improve the efficiency of industrial components product design.The 3D point cloud model is a kind of common data in the digitization of solid model.In order to express the details of the surface of industrial components,the number of points in the point cloud is usually huge.What's more,there is a lot of data redundancy in point cloud.How to extract the information automatically from the original discrete point data through the analysis and mining of models is the crucial content of the model reconstruction in reverse engineering.Nowadays,there are a plenty of researches on extracting feature lines from 3D models.Due to the diversity of industrial components,automatic extraction of line information still requires a lot of manual editing,with degree of automation,accuracy and completeness to be improved.Although most of the researches only aim at feature edges or surface segmentation,in the actual feature extraction process,edges information and patch area information are mutually relevant and affected.In this paper,a complete edge and surface feature extraction algorithm is proposed for industrial components data of 3D point cloud model considering edges and patch area information.In this paper,first of all,we extract the initial feature points of the 3D point cloud model.We use the normal vector information of point instead of coordinate information to construct the normal vector distribution matrix for principal component analysis.Moreover,we set the double threshold to divide the points of point cloud into three types:line,plane,and corner.In the extraction of surface feature,the implicit and parameter expression are used.According to the initial feature points,we first divide point cloud into different facets by using region growing method based on distance constraint and then fit each facet's implicit function.we mainly use three common surface models that often used in the industrial components models,that is,plane surface,cylindrical surface and spherical surface.In the process of solving model parameters,the normal information is added into random sample consensus algorithm to fit each facet's implicit expression function.Subsequently,we use the implicit function and distance constraint to extend,merge and segment the facets in the optimization process.Finally,we choose the non-uniform rational B-spline surface(NURBS)to express each facet point cloud and export the results to a CAD readable file format.In the extraction of the feature edge,we refine the initial feature point of edge set and extract the surface boundary point set from the result of feature surface extraction.Afterward,we divide the two point sets into different lines by using region growing method based on distance and consistency constraint.Furthermore,we adopt the principal component analysis method to analyze the characteristics of every edge and combine the two point sets data with corner information to establish the topological relation.Last but not least,we make advantage of the shortest path algorithm in graph theory to reconstruct the curve and output the vector information.In order to verify the feasibility of the algorithm presented in this paper,we use three groups of point cloud data to carry out experiment.We discuss the setting of the relevant parameter thresholds in the algorithm and analyze the results.The experimental results show that the method of extracting the feature edge and surface is feasible and has a high correct rate,which makes the automatic degree of industrial components modeling improved.