Abstract
Virtual reality is a new emerging displine which studies theories and technologies for synthesizing artificial (virtual) environments and to provide relevant interaction experience for participants. Because VR is aimed to simulating the real world situation, the system must respond to the user's action in real time. For visual response, the system usual generates realistic scence by high performance graphics workstations. However, the performance of graphics workstations is limited by the maximun number of polygons being generated. A complex virtual world norrmally exceeds this limit. A set of techniques have been proposed to render complex worlds in real time, and the Level-Of-Detail (LOD) technique is proved to be a very promising and successful one. In this thesis, we propose an algorithm that automatically generate virtual objects of multiple levels of details based on the " vertices fusion" principle. Furthermore, an experimental LOD subsystem based on propsed algorithm is implemented. Empirical tests show that our algorithm can generated different level-of-detail models efficiently. This algorithm is applicable to any polygonal based 3D objects and it can also be used to generated a sequence of objects which form a hierarchical level-of-detail models. Besides, some visual and geometric similarities are discussed, and examples are given to show that our algorithm preserves these similarties among models of different LOD. Finally, a walkthrough system with LOD capability is built. The hand on experience of the system is satisfactory due to the efficentcy of the algorithm and the preserving of visual similarity between objects of adjacent LOD.