Key Techniques Of Complex Object Surface Modeling Based On Point Cloud

The surface reconstruction based on point clouds has a wide range of applications in smart cities,virtual reality,and reverse engineering,but there are still many difficulties in the modeling process.There is a lot of noise in the point clouds data that affect the calculation results of the local characteristics of the point cloud.Meanwhile,the massive point cloud data will affect the efficiency of surface reconstruction.Point clouds lack of topological information and obvious boundary features,and the curvature of the surface changes greatly,a three-dimensional model cannot be constructed by extracting the characteristic line.Taking the point cloud data of complex curved objects as the research object,we study the key technologies in its modeling process in paper.For large-scale noise existing in the original point cloud data,a denoising method based on statistical analysis was studied.And aiming at the limitations of the uniform mesh algorithm in the loss of the details of the main point cloud when simplifying the point cloud,a point cloud simplification method based on the curvature feature constraint is proposed.Firstly,the curvature of the point cloud is calculated by the local surface fitting method,and the curvature threshold is set to divide the point cloud into a flat area and a steep area.On this basis,the uniform mesh method is used to simplify the point cloud.In view of the limitations of the traditional ICP algorithm relying on better initial position,the method of rough registration based on normal features is studied.First,the local surface fitting method is used to estimate the normal line of point cloud data,the FPFH characteristics of the point cloud are calculated on this basis.Then random sampling consistency algorithm is used to perform initial registration of the point cloud.Finally,the KD-Tree is built to speed up the search efficiency of the corresponding point,and the normal vector threshold is set to remove error corresponding points to improve the ICP algorithm.For the problem of low degree of automation and poor accuracy in the reconstruction of point cloud grids,a method of surface reconstruction based on Crust algorithm was studied.First,the Voronoi diagram of the point set is calculated,and Delaunay triangulation is performed on this basis.Then filter the best triangles,the crossover coefficient is obtained by the distance between the centers of two adjacent tetrahedrons,and compared with the initial threshold,a complete triangular mesh is finally constructed.Experimental results show that the denoising method based on statistical analysis and the simplified method based on curvature feature can obtain pure point cloud data;the coarse registration algorithm based on the normal features can provide a better initial position for the point cloud to be registered,and the improved registration algorithm improves the efficiency and accuracy of registration;the Crust algorithm-based surface reconstruction method has high degree of automation,high efficiency,and the generated triangular network has high accuracy.