Research On Simplification Algorithm Of 3D Model Based On Region Segmentation And Learning

Industrial virtual simulation is a key technology to promote the development of the Industrial Internet.Real-time rendering is the main performance bottleneck of this technology,and a simplified method for high-precision 3D mesh models is an effective way to improve real-time rendering performance.At present,most of the mainstream mesh simplification algorithms reduce the storage size of the original model by eliminating redundant data,but this will lead to the problem of missing features and topological deformation in the simplified model.As a result,it cannot meet the needs of accurate simulation of the model in the simulation system.Therefore,in this paper,the research on the 3D model simplification algorithm based on grid region segmentation is carried out.Firstly,the accurate classification of regional feature information is realized by grid region segmentation,then an adaptive region simplification strategy is constructed according to the region classification information,and finally the adaptive grid simplification is realized by combining the feature maintenance operation.The main research work of this paper is as follows:(1)Mesh region segmentation algorithm based on fusion curvatureAiming at the problems of blurred boundaries and low efficiency in existing region segmentation algorithms,this paper proposes a grid region segmentation algorithm based on fusion curvature and partitioned clustering.Firstly,a graph construction method based on fusion curvature features is designed,combining deep curvature information and shape index to construct surface classification descriptors,which are introduced into similarity calculation to improve the quality of dual graphs of grid models;Secondly,the dual graph is regarded as the dataset to be clustered,a K-medoids clustering algorithm framework based on incremental search and K value estimation is constructed,define the sensitivity function to determine the size of the number of clusters K and the position of the initial center point,and according to the replacement times of the cluster center point,the search range is gradually expanded until the final cluster center point is determined,which is used to improve the stability and efficiency of the clustering and realize the precise area segmentation of the grid model.(2)Adaptive mesh simplification algorithm for region classificationAiming at the problems of topological deformation and missing features in existing mesh simplification algorithms,this paper proposes an adaptive mesh simplification algorithm based on sharpness features and region classification.Firstly,an edge shrinking algorithm framework based on vertex sharpness features is innovatively constructed,the average curvature feature difference between a vertex and its neighbors is calculated by bilateral filtering,and is introduced into the edge shrinkage cost by means of exponential harmony,which is used to reduce the shrinkage priority of edges with important features;Then,a dynamic index optimization algorithm framework based on regional classification is designed,the regional optimal index selector is constructed by grid region classification information and supervised learning network,which is used to predict the optimal simplified index of each local region in the process of edge shrinkage;Finally,the edge shrinking algorithm based on vertex sharpness features is implemented on the original model,and the simplification strategy is dynamically adjusted based on the optimal simplification index set,so as to realize the adaptive mesh simplification for region classification.