Research On Projection-Based 3D Model Retrieval

With the proliferation of computer graphics and computer animations, 3D models are becoming as ubiquitous as images and video. Firstly, graphics proc-essing unit have expanded the applications of 3D graphics. Secondly, populariz-ed modeling tools and 3D scanning devices are making acquisition of 3D mo-dels easier and less expensive, creating a large supply of public available 3D data sets.Thus, the need for an automatic and user-friendly content based multimed-ia retrieval system has consequently become urgent.The primary challenge for a content-based 3D model retrieval system is toextract the proper features for representing the diverse shape of 3D models forefficient indexing of similar 3D models.Techniques for effective and efficient content-based retrieval of 3D modelshave therefore become an essential research topic. In this paper, a novel feature, called -based is proposed for 3D model retrieval. The descriptor is invariant to translation and scaling of 3D models and it is robust for rotation. An efficient similarity matching method is used to find the best match for an input model.The experimental results show that the proposed retrieval methods are effective and efficient.