Fast Rendering of Massive Textured Terrain Data

Rendering of textured terrain models has become a widely used technique in the field of GIS applications and virtual reality. In this paper, we propose a framework based on tile-pyramid model and linear quadtree tile-index, which enable the real-time rendering of out-of-core terrain data sets while guaranteeing geometric and texture accuracy. The digital elevation model tile pyramid and the orthophoto tile pyramid are used to render textured terrain without real time resampling the data. The quadtree is used for constructing tile-index and managing data tiles. The process of constructing linear quadtree based on pyramid is divided into three steps: logic partitioning, node coding and data partitioning. The real-time loading of terrain dataset is achieved with the view frustum culling technology and the target-tiles searching algorithm based on resolution-testing and the tile-request prediction mechanism. The dynamic bintree triangulation is used for the real-time rendering of the terrain tiles. The experiment results show that, based on the current personal computer, this framework can achieve good performance for real-time rendering of massive terrain dataset whose size is unlimited. We have shown that off-the-shelf hardware is powerful enough to render huge textured terrain datasets, and are eager to explore the rendering capabilities of our framework with even larger and more detailed datasets.