Content-based 3D Model Retrieval

With the recent development of multimedia technology,computer vision and 3-D graphics hardware, the number of 3d models are growing rapidly, and 3d models are more and more widely useful in many fields such as CAD/CAM and computer animation. Now a large number of 3d models are available from internet. So there is an increasing need for a search engine to help people quickly find the models that they need. This paper focus on content based model retrieval. Based on the up-to-date methods and technologies for content-based 3D model retrieval, two novel feature extraction methods are put forward in this thesis: the shape descriptors based on U system rotation invariant moment and based on V system rotation invariant moment.In this thesis, we review the development of the 3d models retrieval system; discuss the advantages and disadvantages of these primary techniques, and then present model retrieval methods based on orthogonal polynomial moments and a detailed project is proposed to testify the algorithm. First, we need to do some pro-processing for the general mesh model in order to obtain shape descriptors that are invariant to the translation and scaling. Second, the ray casting technique is used to extract geometry feature of 3d models. Third, we construct shape descriptors based on orthogonal polynomial moments that we proposed and use them in model retrieval experiments.