Grid-based LOD Algorithm Optimize Research For Terrain

For a long time past, all of the interpretation results from GIS data have been displayed or mapped by having resort to 2D and static approach. That creates difficulty for people to tell immediately and positively what the overall results means about the terrain surface. Also, the relevant information obtained could not be brought into its full play. The technique of visualization in scientific computing paves a way to rebuild 3D terrain model using the interpreted result of GIS data.This thesis focuses on the theoretical and methodological studies in the several subtopics: 3D terrain visualization and rendering, terrain simplification, LOD(Level of Detail), and multiresolution terrain representation. These studies are based on lots of advanced results of 3D computer graphics, scientific computation visualization, and virtual reality. A quadtree-based optimization algorithm for dynamically generation continuous levels of a given complex landscape in real time is developed. In this algorithm, an thought of "partitioning the terrain both spatially and resolution" is applied. It first partitions the large DEM data into some smaller sub-blocks and then stores the data in resolution order in each sub-block, the view-dependent dynamic node error evaluation method is also proposed. Rendering speed is greatly improved by using only one quadtree traversal in mesh generation.At last, based on the researched result of the above problematic situation, the three-dimensional visualization system has been established combining with experience and practice in the item of "3D-Dynamic Simulation System (derivating the water from Changjiang to Taihu Lake)". The system has these functions: Data can be received and pretreatment, three-dimensional visualization and dynamical display of terrain, three-dimensional model real-time dynamic simplification and multiresolution representation. Empirical tests show the method is robust, and fast.