Research On Feature-driven 3D Model Resampling Methods

In the field of education technology,computer aided teaching system is popular among the masses of teachers and students.Triangular mesh model as the foundation of information expression carrier,has frequently applied in the computer aided teaching system,at the same time,as the fourth kind of media form,has widely applied in 3D animation,3D games and industrial manufacturing.The resampling is beneficial for 3D mesh to improve its quality,and to make it better suit for various applications.In this paper,compared with the conventional resampling algorithm,our algorithm adds user editing in the process of resampling,allows users interactively control the sampling distribution,and generates the new 3D model including user information.Meanwhile,the distribution of the new model's vertices has the characteristics of local adaptability.The work of this paper mainly includes the following several aspects:1.Planar parameterization is important basic operation in resampling algorithm.To keep the plane parameterization results of boundary continuity,we select global parameterization method.Discrete Ricci flow plane parameterization has a better conformal,can reduce the deformation in the process of parameterization in a certain degree.The method using inverse distance CP,defines Riemannian metric on the vertices of grid model,through the Gauss-Newton method,and make it induced the gaussian curvature of consistent with the preset target curvature.At this process,the 3D model can unfold onto the plane.2.We propose a uniform resampling method based on the plane parameterization algorithm.In this algorithm,we use the shortest path algorithm to cut a closed 3D model.After that we apply the above planar parameterization method to open mesh to parameterized operation.In order to make the sampling point set distributed uniformly in three-dimensional space resampling,and twodimensional space should be sampling point set based on parameterized area in the process of deformation measurement of the density and orderly distribution.Deformation in the area of large area of intensive sampling point set distribution,deformation in the area of the smaller regional distribution sparse sampling point set,so we need to calculate the area in the process of parameterized deformation,and define it as the two-dimensional spatial sampling density function.Finally,we transform 2D sampling point set for the triangular mesh,and its reconstruction in 3D space,get 3D point set uniform distribution of the new 3D mesh.In this paper,our sampling algorithm can modify the density function,to control the density of sampling points.3.After the uniform resampling method,we propose a feature-driven adaptive resampling method.In this method,how to make the sampling point set according to the feature of the model itself adaptive distribution is an important issue.We use the Centroidal Voronoi Tessellation to implement adaptive sampling point set.After the above planar parameterization operations,we construct geometrical image gray level through geometric metrics of the original grid and deformation metrics in the process of planar parameterization.Then we extract the edge feature line of image,and add the content of user editing,after that according to the image include features line to defining sampling density function.Centroidal Voronoi Tessellation use the basic principle of the Voronoi diagram,according to the density function,divide the sample point area,and get the two-dimensional space adaptive sampling grid.This article has carried on the experimental results show and analysis,after the narrative of the basic idea in the above resampling algorithm.The experimental results mainly display as,different sampling ways of different models,different density distribution and the user edit.