Research And Improvement Of The Algorithms For Real-time Terrain Rendering

Terrain rendering is one of the core technologies in constructing Three-dimensional(3d) GIS, which use large amount of triangles to represent the terrain surface's geometry and render these triangles on screen to get 3d visualization effect. At present, the scales of the terrain data is getting larger and larger and the resolution of the terrain data is getting higher and higher, which have exceeded the rendering ability of current computer graphics cards. So the terrain rendering has to face tough challenges. To solve this problem, this paper discusses it from the software aspects and makes a detail analysis of the CPU oriented algorithms for example ROAM, Quad tree, in the split method, simplification rules and visual artifacts aspects. The paper also introduces the GPU oriented algorithms, for example Chunk LOD, GeoMipMap, which partition the original terrain into patches at different resolution. Since there are some shortcomings in both the CPU oriented algorithms and GPU oriented algorithms, this paper provides specific solutions to improve them and puts forward a new algorithm based on the seamless division of terrain to implement large scale interactive terrain rendering. The main achievements are listed as following:1. Utilizing special terrain division method to fix cracks existing between two adjacent terrain patches without any extra computations and graphics rendering.2. Adjusting the terrain data structures so that the dynamic construction of terrain triangulation is committed during data preprocessing stage other than runtime rendering stage.3. Creating mipmap for textures manually and dynamically to avoid aliasing artifacts and save system storages.4. Improving the schemes of terrain data loading and unloading by an independent thread which complete to load all the terrain data during several frames and reduce the waiting time for GPU to get terrain data.5. Changing the rendering primitive type from vertex or triangle to triangle-strip which can get best performance of current computer graphics cards and increase the triangle throughput and avoid repeating computation for the same vertices. 6. Utilizing the programmable features of GPU to write some program, which will be running on GPU, by high level shader language so that it can reduce the workload of CPU and benefit from the float computation performance of GPU.