Multi-Resolution Model Algorithm For Large-Scale Terrain

Visualization of terrain information is playing a key role in many applications such as geographic information system, simulation system, visual reality and 3D games, etc. With the rapid development of the remote sense, satellite technology and computer technology, people expect to investigate or view larger terrain regions with finer precision, which inevitably leads to the real-time rendering of huge terrain dataset. This thesis explores the strategies of real-time rendering for large-scale terrain, and the principal content in this thesis is as follow: Some application backgrounds of the real-time rendering of large-scale terrain data set are introduced. It is unveiled that the nature of the real-time rendering for large-scale terrain is to find out such rendering algorithms that have no or little dependency with the number of the triangles in scene. And it is pointed out that there are three fundamental strategies to accelerate the rendering procedure as visible-surface detection, level of detail and image-based rendering. Some resolutions which are presented in recent publications about terrain rendering are summarized. The characteristics and classes of level of detail are described, as well as a series of related issues. The application with level of detail in large-scale terrain rendering is discussed. And a series of algorithms about terrain multi-resolution modeling in recent years is enumerated and discussed. In this thesis, a system for large-scale terrain real-time visualization is proposed. Within this system, the approach integrating multi-resolution representations of terrain and fast view frustum culling is presented. A novel method for real-time multi-resolution modeling is presented. In order to guarantee the precision of the multi-resolution model effectively, the conservative estimation of the projected error in screen-space is used as the criterion which is used to choose an appropriate level of detail, which can be stored and computed efficiently. Furthermore, a fast scanning is used for view frustum culling based on the projection of the view frustum on x-z plane. To...