MULTI-RESOLUTION REPRESENTATION OF DIGITAL TERRAIN AND BUILDING MODELS

This research develops effective algorithms for multi-resolution representations of three-dimensional (3D) digital terrain and building models to achieve better performance in cyber city applications. The objective is to create multiple levels of detail (LOD) of terrain meshes and polyhedral building models so they can be used efficiently according to viewing parameters and application requirements, while preserving critical features of the datasets. For terrain meshes, a tile-based approach is adopted. A mesh refinement algorithm based on modified quad-tree process is developed to generate multi-resolution representations of each terrain patch. On the other hand, a divide-and-conquer strategy is employed for the generalization of 3D building models to formulate LOD representations of complicated buildings. The idea is to apply generalization in 2D orthographic views of original polyhedral building models and then reconstitute simplified 3D models accordingly. Experimental results with complicated terrain and building datasets demonstrate that the developed LOD algorithms can improve cyber city performance significantly.