Research On 3D Mesh Clustering Segmentation And Retrieval Method Based On Salient Points

In recent years, with the rapid development of Internet and the three dimensional animation technology. At the same time mesh model visualization, construction and reconstruction of corresponding technology are in continuous progress, data amount and the complexity of 3D model of rapid growth, the computer real-time graphics processing faced with a severe test. Therefore, the mesh segmentation problem is proposed, which can greatly reduce the complexity of the three-dimensional model, thus making it become the focus of many research and practical workers.Due to the complexity of the 3D model, uncertainty and different segmentation results demand, the segmentation of 3D model has not a unified algorithm and most of the algorithms are obvious defect segmentation. Also feature extraction algorithm of 3D model is always in the improved, but algorithm search results are not very good, and not a general feature extraction algorithm can for all the models have the best retrieval results, and average retrieval better algorithms for the effect of some model may be very poor. Feature extraction and representation of 3D model is the key to the research of 3D model retrieval based on content. Each method can reflect and measure the information of the 3D model in a certain aspect, which has a certain performance and advantage. But so far, there is no way to fully reflect the human shape of the recognition and understanding. Based on this, this paper proposes an improved fuzzy clustering method to solve the above problems. The specific research contents are as follows.(1) 3D model retrieval related research. Firstly, it analyzes the 3D mesh segmentation and traditional segmentation methods, analyzes the application of clustering algorithm in the segmentation of 3D mesh model, finally the 3D model retrieval architecture and retrieval method and the three dimensional model of information retrieval application system has carried on the thorough analysis, laying the foundation for the next further work theory improvement and prototype system design.(2) This paper proposed a salient-point detection algorithm based on curvature flow, This method combines the classic Mesh- Do G and Salient Point feature point extraction method, The experiment found that these significant points can be a good marker model of regional location information, can achieve better segmentation.(3) During the retrieval phase, this paper conducted the retrieval of mesh model by applying the global features of whole-part integration. Considering the various deformations of the model, the feature description selected in this paper was HKS(heat kernel signature) by equidistance, isometry and invariance; Binding characteristics to achieve the overall and partial retrieval model. Final results of the analysis and comparison.