Large-scale terrain visualization system in a cluster environment

Abstract

In this paper we propose an efficient Large-scale terrain visualization system in a cluster environment. Firstly, we implement level-of-detail (LOD) building in parallel, with an obvious time saving compared with a single-machine environment. The LODs are then stored in the server. While rendering, clients get LODs on demand. LODs are organized with a scene graph structure, thus view frustum culling can be performed efficiently. A data thread is introduced to particularly manage data exchanged with the scene graph, cooperating with the rendering thread. In addition, we use GPU optimized terrain rendering algorithms to achieve high triangles throughput. To summarize, with our system, original large-scale terrain data can be built into LODs in a relatively short time and LODs will be rendered interactively.