大规模地形实时绘制算法

该文提出一种适合大规模地形实时绘制的简单高效的LOD简化算法。该算法使用一种紧凑有效的规则网格表示方法，优化网格节点的数目，减少可视化过程中的计算量，降低额外内存开销。探讨该算法相关的数据组织、视域裁剪、LOD层次选择、裂缝消除、三角形化等关键问题。实验结果表明，该算法实现简单，内存开销较少，CPU耗费小，对图形卡要求低，能够在普通机器上实现大规模地形的实时漫游。

A Simple and Efficient Algorithm for Real- Time Rendering of Large Scale Terrain

Large scale terrain roaming has been widely used in games, simulation, virtual reality, etc.There terrain scenes may contain thousands of polygons and although modern graphics cards allow the display of many thousands of polygons at real-time framerates, many applications have models with geometric complexities that, by far, exceed the real-time capabilities.In order to reduce the number of polygons to be rendered and thus achieve real-time framerate many research papers have dealt with different level of detail (LOD) algorithms and aggressive frustum culling.However, many current terrain simplification algorithms cost large memory and need much real-time computation in visualization.It is difficult to build large scale virtual terrain environment and to achieve real-time framerate by means of these algorithms.Based on the characteristics analyses of the current terrain simplification algorithms, this paper proposes a view-dependent refinement algorithm for continuous LOD in terrain block.The algorithm uses a compact and efficient regular grid representation and keeps terrain data into memory unchangeable.The whole terrain scene in memory is divided into many tiles and every tile has the same level of detail and is taken as one node.The method has the advantages of the optimization of the grid node number, reduction of computation, less utilization of extra memory.Several key problems such as data organization, frustum culling, choosing LOD level, avoiding cracks and triangulation are also discussed.Finally, the paper designs a prototype system and accomplishes simple implementation using a case study of the Yangtze River region, China.The terrain size of the region is (20 047)*33 287.Experimental results show that the algorithm implementation becomes very easy, and its efficiency in computation and memory is high and requires low-end 3D graphics card.The algorithm also can support the real-time roaming for large scale terrain in general desktop computers.