一种基于现代GPU的大地形可视化算法  

GPU Friendly Terrain Rendering Algorithm

作　　者：潘宏伟[1] 李辉[1] 廖昌阊[1] 曾安祥[1] 

机构地区：[1]四川大学计算机学院图形图像研究所,四川成都610064

出　　处：《系统仿真学报》2007年第14期3241-3244,3275,共5页Journal of System Simulation

基　　金：国家自然科学基金项目(60303034)

摘　　要：地形渲染在计算机游戏,飞行模拟和视景仿真等领域的应用越来越广泛,随着渲染场景复杂度的增加,每次需要绘制的地形数量也越来越庞大。同时,新一代的显卡绘制能力的不断增强,原有的许多地形渲染算法已经不能很好的满足用户需求。在总结现有算法的基础上,提出了一种基于现代GPU的地形渲染算法。该算法同样使用高程图作为地形数据,将地形分成很多小块,每次渲染时以块为单位,所有小块使用四叉树组织成一个层次化结构,不同层次的节点代表了不同细节层次的地形范围,并且采用了与Mipmap类似的细节简化方式,渲染时不需要对分块重新简化。为了保证CPU和GPU的负载平衡,将一些复用率高的地形分块缓存到显卡中,大大降低带宽需求。实验证明该算法可以更为充分利用图形处理器的加速能力,既能满足渲染精度要求也能达到一个较高的帧率。Terrain rendering is more and more used in many fields such as computer games, fly simulators and scene emulator applications. With the increasing of complexity of the scene, the amount of data to be rendered becomes larger .At the same time, the modem graphics card has a more powerful rendering ability, so many traditional terrain rendering algorithms can not fulfill the requirements. A new algorithm was proposed that was friendly for modern graphics card. The algorithm also used the regular height map grid as terrain data. Terrain data was tiled and organized by quad tree. Each tile covered a part of terrain with different detail. The method used which Mipmap used to refine the mesh, and there was no remising in real time rendering. To balance the load between CPU and GPU, many tiles used at a high frequency were cached in GPU, which could reduce the bandwidth requirements significantly. The results show the algorithm which is well-suited for recent GPU can not only meet the requirements of detail but also reach a higher frame rate.

关 键 词：地形可视化 模型细节层次 几何缓存 GPU VBO 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]