An Effective Visualization Method for Large-scale Terrain Dataset

-Visualization of very large digital elevation models plays an important role in many applications ranging from computer games to professional Geographic Information System (GIS). However, the size of terrain datasets is much larger than the capacity of main memory, which brings up serious challenges for the interactive visualization of massive terrain. In this paper, an effective visualization method for large-scale terrain is proposed. In preprecessing stage, the full resolution terrain dataset is divided into several uniform blocks. For each block, a LOD hierarchy, which contains different resolution of elevation data, is constructed based on quadtree-split. In order to save the external memory, a non-redundant storage method is put forward. Different from the traditional pyramid model, we stored only the newly added vertices in each level. The data quantity during real-time rendering decreased effectively due to the sight-line-dependent screen error computation, view frustum culling and detailed culling scheme. To ensure the continuity of navigation, a preloading scheme, which based on the view frustum extension, is presented and implemented. Experiments show that compared with traditional methods, our method has reduced the external memory cost by 60% and achieved good visual effect with high rendering speed. Key-Words: Levels of Detail (LOD), Model Error, Static Error, Multi-resolution model, Quadtree