在GPU上实现地形渲染的自适应算法  

Adaptive Refinement for Terrain Rendering on GPU

作　　者：王旭[1] 杨新[1] 王志铭[1] 

机构地区：[1]上海交通大学图像处理与模式识别研究所,上海200240

出　　处：《计算机辅助设计与图形学学报》2010年第10期1741-1749,1755,共10页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(60572154)

摘　　要：为了满足飞行状态实时监控系统对地形渲染快速逼真的要求,提出一种基于GPU的交互式地形自适应渲染算法.该算法中每帧渲染包含2遍GPU处理过程:第1遍采用流计算的方式,利用固定网格映射方法生成粗糙地形采样网格,在此基础上,根据地形表面复杂度计算粗糙采样网格中每个三角形的细化深度;第2遍进行地形的渲染,根据第1遍计算出来的每个面片的细化等级,选择初始化阶段预存储在GPU缓存中的不同细化等级的网格模板,对粗糙采样网格进行自适应细分,为了消除T型连接的问题,在顶点着色器中包含了一个网格模板的匹配处理过程.最后采用高程数据配合卫星照片的方式,生成具有高度仿真的三维虚拟地貌.基于文中算法实现的实时监控系统在支线飞机的飞行试验中取得了较好的效果.In order to meet the requirement of more realistic and efficient rendering in a flight monitoring system,we present an interactively adaptive terrain rendering algorithm on GPU in this paper.By the algorithm,a two-pass GPU acceleration technique is designed for each frame.In the first pass,stream computing is applied to a coarse sampled mesh generated by persistent grid mapping.Then the refinement level for each triangle in the coarse mesh is calculated based on the complexity of the surface topography.In the second pass,terrain rendering is made according to the refinement level of each triangle,and the coarse sampled mesh is tessellated using different mesh refinement patterns which are updated and stored in the vertex buffers in the initialization stage.In order to eliminate the problem of T-junction,a matching process for various refinement patterns is implemented in the vertex shader.At last,elevation data and the corresponding satellite images are combined together to generate more realistic terrain in real time.The algorithm proposed has been applied to a flight monitoring system to support the flight test of advanced regional jet.

关 键 词：GPU 流计算 固定网格投射 自适应细分 细化模板 实时渲染 

分 类 号：TP931[自动化与计算机技术]