3D Model Retrieval Based On Both The Shape Feature And MRG Skeleton

With the development of computer technology and the widespread application of three-dimensional model, how to research desired three-dimensional model from a large database quickly and efficiently need people to address the important research projects as soon as possible. Three-dimensional model retrieval based on keywords is a kind of early technology, this technology has been developed for a long period of time. The model retrieval based on content technology is a kind of new technology, which has its own advantage, it overcomes the deficiencies of traditional model of retrieval. The use of three-dimensional model characteristics of the various aspects of indexing to achieve retrieval, content-based retrieval model technology is more objective and efficient.3D retrieval methods based on text usually lack in meaningful description for 3D models, because the text that people add to 3D model is subjective. So this method can not describe 3D model objectively and add some difficulties to 3D retrieval methods based on model content can describe some model feature, these features describe model's topological structure and geometrical shape .3D retrieval methods based on model content remedy a defect of retrieval methods based on text, it have became important content in 3D retrieval method research.According to multi-resolution Reeb graph method (MRG), we propose an optimized articulation-constrained skeleton extraction approach. The multi-resolution Reeb graph method overcome the low efficiency ,this method must calculate discrete contour constrictions of each vertex on 3D model surface, and it can instead obtain the hyperbolic extrema points as constrained points .These constrained points imply the concave feature of the local surface, and the articulation feature contours can be then extracted. The Reeb graph is finally improved by adding these new feature points. The 3D model skeleton keeps well topological structure and local features of 3D mode.Through studying other techniques, we propose a new 3D retrieval method .This 3D retrieval method research model from model database based on MRG skeleton tree structure. We extract the Reeb graph skeleton based on Multi-resolution Reeb graph (MRG)theory which indicates the model topological structure is generated. It transforms Reeb graph of 3D model to a tree structure. The skeleton of Reeb graph can account for only topology features, but can not compare the shape detail of model at a local level. The method acquires the geometrical features of 3D model by computing the discrete curvatures and areas of corresponding segments for every critical point in model skeleton. Combining the topology structure and local geometric details of 3D model, we propose an improved similarity and difference functions. The experimental results of our method have shown the robustness and high efficiency of this method in computing 3D models similarity.Two 3D retrieval methods based on MRG skeleton tree structure are proposed. We count the number of children in skeleton tree structure as topological feature of skeleton node. Improved 3D model Retrieval based on MRG skeleton count discrete curvature, and get summation of discrete normalized curvature as geometrical feature of skeleton node. 3D model retrieval based on both the shape feature and MRG Skeleton get the area rate of local area and whole model, this area rate can describe the importance of local area. we count curvature variance of local area to describe model local shape feature. A series of experimental results have shown its accuracy and efficiency.