| Summary: | 碩士 === 中華大學 === 資訊工程學系碩士班 === 94 === Abstract
The advances in 3D data acquisition, graphics hardware, and 3D data modeling and visualizing techniques have led to the proliferation of 3D models. The searching for specific 3D models becomes an important issue. Techniques for effective and efficient 3D Model content-based retrieval have therefore become an essential research topic. In this thesis, two novel feature, called adaptive elevation descriptor (AED), and grid sphere descriptor (GSD), are proposed for 3D model retrieval. The adaptive elevation descriptor (AED) is invariant to translation and scaling of 3D models and it is robust for rotation. First, six elevations are obtained to describe the altitude information of a 3D model from six different views. Each elevation is represented by a gray-level image. Next, the MPEG-7 Angular Radial Transform (ART) is applied to six elevations to extract feature vectors. Since, there are six elevations for each 3D model. An efficient similarity matching method is used to find the best match between two models. The grid sphere descriptor (GSD) is also invariant to rotation, translation, and scaling of 3D models, and it can give the interior characteristic the 3D model. Experimental results show the combination of AED and GSD is superior to other descriptors.
|
|---|
