Real-time Rendering Of Large-Scale Terrain Based On Wavelet Analysis And GPU

3D Terrain which is the basis of virtual environment have been brought into wide use in relative field. The demand of terrain size is more and more wide because of the GIS and Digital-Globe are more widespread, so the lage-scale terrain contained millionaired triangles is common today. Such massive terrain data is inconsistent with the real-time rendering because of the ability of hardware has its limit, and therefore the simplification and multi-resolution become an essential procedure of rendering. The thesis has a research on multi-resolution of DEM data based on wavelet analysis and GPU, and put forward the proposal in relative problems. In detail, the main contributions of this thesis are as follows:●In the thesis, We propose a multi-resolution approach for large-scale terrain which based on wavelet analysis, choose the appropriate wavelet function according to the characteristic of terrain simplifying, generate the terrain grid by wavelet analysis, add in high frequency wavelet coefficients bit by bit to achieve the continuous level of detail.●To point against distortion of boundary caused by wavelet analysis, we expand the range of sampling, group the terrain chunks as a whole unit in wavelet analysis, so as to eliminate the distortion of boundary effectively.●It will generate crack between different chunks of different resolution because of thier different quantity of vertexes. After sum up extant approaches, we propose an approach by translate the vertexes in boundary, stitch the crack effectually. The algorithm can also fit the crack made by distortion of boundary between chunks of same resolution, and don't need increase or delete vetexes, so it will be highly efficient.●To improve the efficiency of occlusion culling, we present an algorithm for occlusion culling based on GPU, set up a queue for saving the searched result to point against leisure time of CPU and GPU.●We transferred the computing of wavelet analysis to GPU, because the quantity of computing is vast. We present an algorithm of wavelet analysis based on matrix and texture, and decrease the time of computing greatly.