Research On Spectral Clustering Method For Complex Network Of 3D Model Adjacency Surface

With the continuous deepening of computer information,the aided design and other industries are developing rapidly.Manufacturing enterprises have accumulated a large number of 3D CAD models,condense the design results and wisdom of people.With the gradual maturity of 3D CAD technology,it is facing the new challenge of large-scale model library processing.How to make full use of model database,studying the clustering analysis and retrieval application of 3D model has become an important research topic in recent years.At the same time,there are many feature description methods and data storage formats of 3D models,which can be presented in the form of complex network of adjacent surfaces.The development of complex networks and their applications in different fields can help people find the desired information.In the field of 3D model,the clustering method can be used to excavate the structure of the complex network of adjacency surfaces of 3D model,so as to realize the clustering of 3D model nodes and surface nodes.Therefore,the main work of this paper is as follows:(1)A multi-dimensional feature modeling method based on STEP file is proposed to complete feature extraction of 3d model.The complex network of adjacency surfaces of the 3d model is established by using the element information of STEP file decomposition.Global and local feature vectors are constructed,including geometric feature vectors,topological feature vectors,conformal descriptor vectors,and core surface boundary feature vectors,and STEP files of 3D models are mapped to the mathematical model of multidimensional feature vectors.(2)The spectral clustering algorithm based on graph theory is applied to the clustering analysis between 3D model nodes and plane nodes in this paper.The experimental results show that the spectral clustering algorithm can realize the clustering analysis of 3D model nodes and surface nodes.The comparative experiments show that the proposed algorithm is more effective than the traditional clustering algorithm.(3)The 3D model fusion feature similarity measurement model is proposed.Firstly,the similarity measures of the feature vectors obtained by different modeling methods are carried out re-spectively.Then,the weights corresponding to different features are calculated according to information entropy.Finally,the similarity of different features is weighted to obtain the final similarity measurement results.A new similarity calculation method is applied to the spectral clustering algorithm of 3D model instead of the traditional similarity measurement method.At the same time,it proves the effectiveness of using the Lanczos method to obtain the eigenvalues and eigenvectors of Laplace matrix.Using the eigenvalue gap to determine the number of optimal clustering categories,the spectral clustering analysis between 3d model nodes and surface nodes is completed in the corresponding feature vector space.complete the clustering analysis between nodes in the corresponding eigenvector space.(4)A retrieval method for adjacency surface complex networks based on clustering results.Based on the clustering results of 3D models,the retrieval experiment is carried out.Through comparative experiments,the model retrieval method presented in this paper shows high accuracy and efficiency,which proves the effectiveness of the proposed retrieval algorithm.