| With the rapid development of computer technology, computer simulation, flight simulation, and other aspects make very high demands on terrain rendering. On the one hand, increasing complexity and growing great data of scenes, on the other hand, the requirements for effects and speed of terrain rendering do not decrease,but increase. Although the current graphics hardware is capable of rendering mass triangles, but geometric throughput of the hardware is still a bottleneck.This thesis mainly uses two technology to overcome these bottlenecks of graphic hardware. One is visibility-driven terrain acceleration technology; the other is continuous level-of-detail technology. Visibility-driven terrain acceleration technology, in the course of terrain roaming, computes potential visible triangle-patches set in real-time, and get rid of these triangle-patches before rendering. As in current majority continuous level-of-detail algorithms, data rendered to graphic hardware is not optimized, and need to reconstruct terrain grids every frame. For the shortage of these algorithms, we put forward an effective continuous level-of-detail algorithm, which preserves unchanged data to greatest extent, render changed data in optimal format to graphic card in one time, which will decreases the data bandwidth. Secondly, we adopt frame delay calculation to solve the lack of view-sensitivity, which many previous view-dependent LOD algorithms have. In addition, we adopt effective memory management to improve algorithm performance. In the process of real-time terrain roaming, nodes data need efficient management, we adopt two memory pools to achieve efficient management: a static memory pool, allot all the memory space on the initialization; A dynamic memory pool, which is a chain, points all the vacant space within the used memory space. This algorithm, based on the Microsoft DirectX graphic library, has been realized in common PC and gets a preferable vision effect and roaming frame speed. |
PDF Full Text Request