Research And Implementation Of A Distributed System Of Large-scale Spherical Terrain

The latitude and longitude coordinate system is a widely used geographiccoordinate system, It is natural to think of the grid can be split in accordance with thelatitude and longitude coordinates when building spherical terrain model. However,because of the latitude and longitude coordinates are determined based on the angle,resulting in a huge difference on area of different grids in the same level, and eventuallylead to significant data redundancy.To solve this problem, this paper proposes a regular hexahedron split algorithm.The first step is execute quad split on each face of the regular hexahedron, and then thegrid should be projected onto the regular hexahedron’s external ball, after that, we getthe sphere grids. Although the problem of differences on area still exists, the four arcs ofthe grid are the same length, so the area problem will be somehow reduced. Make fulluse of the self-similarity between the split grid with the original grid, and the symmetryof the six faces of a regular hexahedron can greatly simplify operations, improve theefficiency of the system.To deal with the height and texture data of large-scale spherical terrain, we use splitalgorithm to organize them into a pyramid structure, and then use compressionalgorithm to pack them together. For data scheduling management, we build adistributed data server, and develop a cache algorithm in the server, in order to achievefast scheduling and sharing of data.The main feature of our rendering system is cross-platform, and even mobileplatforms, which developed fast in recent years. The main step of rendering is updatingthe quad tree; there are lots of work to do in the updating step, including frustum culling,screen error calculation, and mult-resolution model generation. After all these are done,we use GPU-accelerated OpenGL interface to complete the draw.