The design of content-based 3D object model retrieval systems is emerging as a key technology toward interactive multimedia service in the trend of new millennium such as education tools, manufacturing processes, mechanical CAD model databases, medical data analysis applications, archeological model reconstructions and virtual reality simulation. The problem lies in the nature of three-dimensional shape, which is difficult to store and represent as opposed to two-dimensional shape.;This research aims to extract features that describe properties of the 3D object model. These features are indexed and subsequently compared to one another during query processing to find similar 3D object models in the database. This research proposes a new 3D shape descriptor together with 2D multi-view descriptors, which are combinations of existing shape descriptors, to provide a robust 3D object model retrieval system. This new paradigm is to solve complexity in the traditional method of 3D object shape descriptors and lack of robustness in query methods of 3D object model databases.;The first discussion is used the existing 2D and 3D shape descriptors and object representation schemes by which types of feature can be determined. Then, some novel shape description methods are introduced, which can be used for 3D object model indexing and solution for both 2D and 3D input queries. In fact, there exists a few content-based systems for managing 3D object model database. 3D object recognition systems focus mainly on developing sound representation schemes that allow extraction of useful 3D information from range image because their main purpose is to find objects of interest from particular scenes. On the other hand, this research provides an efficient retrieval method based on 2D and 3D shape similarity of the 3D object model where a prevalent existing CAD representation scheme can be used as it is.
展开▼