Semantic-based 3D Model Retrieval

As the development of 3D model capturing technology, 3Dgraphics modeling method and graphics hardware technology, many 3D model libraries have been established. How to browse and retrieval 3D model efficiently becomes a researching hotspot. However, most of the existing retrieval methods are based on low-level features, and low-level features are meaningless to users. There is a semantic gap between low-level features and high-level semantic, so the performance of the traditional 3D model retrieval methods is always disappointing. Semantic based 3D model retrieval methods are more appropriate for human.In this paper, we compare the differences between semantic based retrieval and traditional retrieval methods, and introduce the history and principle of semantic based image retrieval. We also analyzed the why semantic based methods can improve the retrieval performance, and feasibility of semantic based 3D model retrieval. Because the existing semantic based 3D model retrieval methods are not very satisfying, we want to design a simple and practical 3D model semantic retrieval framework based on the existing image semantic retrieval methods.We also design and implement a experimental system according to the framework we designed, and the experimental result shows that the framework is promising in the aspect of 3D model semantic detection.