3D Model Decomposition Based On Fuzzy Clustering And Its Application

With the development of software and hardware of the computer, inrecent years, the way in which the figure of the 3D model is being countedwith geometry increases, and the application of the 3D model is extensiveday by day too, for instance model designing , virtual reality , 3D game ,simulation emulation of the industrial products etc. Today reconstructingmodel have become a much important way, how put existing mesh in anposition, how to modify the existing model according to the new designobject, have already become an important problem. So the meshsegmentation becomes a hot subject for research in recent years.According to the geometry and topological characteristic, 3D modelmesh segmentation (abbreviate as the mesh segmentation) decompose themesh into simple sub-meshes mean, or a directional one shape has shedwhile being 3D close net polyhedron, according to its surface geometry,topological characteristic, divide into the limited, each work with simpleform meaning, and each sub net scene that feed through of a group of figure.Mesh decomposition benefits many applications. In metamorphosis,compression, simplification, 3D shape retrieval, collision detection, texturemapping and so on.In this paper we use fuzzy clustering and cut technology of Binarydecomposition. The algorithm proceeds from coarse to fine. A key idea ofthe algorithm is to first find the meaningful components, while keeping theboundaries between the components fuzzy. Then the algorithm focuses onthe small fuzzy areas and finds the exact boundaries which go along thefeatures of the object. To find fuzzy components, the condition is that everyface should belong to exactly one patch, and allow fuzzy membership.During the decomposition, we increase an mesh pro-processing, insure themesh is connected each Binary decomposition.In 3D model retrieval system, the query methods based on textkeywords, 2D sketching, model matching, and iterative refinement.Presently there isn't part model query method. If user wants to retrieval apart model, he must search model first, and then cut or segmentation thepart model from the whole model, this process is much complexity. Tosatisfy the need of the user, this paper adds part model query method in 3Dmodel retrieval system. We choose 57 models from Princeton ShapeBenchmark, decompose them using fuzzy clustering and cuts algorithm, andget 663 part models, then get the shape descriptor of the part models, adoptcluster's technology, realize to search part models, adopt ClusterTreealgorithm to establish part model database, improve the searching efficiency.Here are the main stages: (1) Choose models from Princeton ShapeBenchmark and cut the models and get the meaningful components;(2) Getthe shape descriptors of the part models;(3) Adopt clustering technology;adjust the clustering result, then use ClusterTree algorithm to establish partmodel database index structure.