Research On Data Management And Transmission Technology Of Massive Terrain Scene

Scene rendering of massive terrain is the basis of spatial representation. It is crucial part of most virtual reality system, such as GIS or virtual battle field etc. High resolution data are obtained from all kinds of sensors. It provides terrain rendering more realistic effect. Meanwhile, quantity of data increases constantly.To ensure real-time rendering of massive terrain, data must be managed, retrieved and transmitted with high efficiency. To satisfy those requirements, this thesis researches several key technologies of data organization and transmission for massive terrain scene. These technologies include design of scene data organization model; efficiently retrieval of massive data; P2 P based distributed storage management of massive data; and compression and transmission technologies of terrain data. The main works and contributions of this thesis are stated as below.(1) To satisfy the requirement of management massive data in virtual battlefield, a context based ontology model is present. Through conceptual abstraction, the ontology model can achieve data uniform representation among subsystems, as well as data sharing and reusing. Using Context limit, the ontology model can reduce data set to be retrieved and aggregate data to be rendered. This can improve the performance of rendering system. A layer judge algorithm is designed, which can support concept adding and ontology model extension. Initial experiments show that the model can satisfy the requirements of uniform representation, expansibility. It also supports efficiently retrieve of data, and parallel operate on data for virtual battlefield.(2) A mid-point index coding mechanism and a data retrieving algorithm using viewpoint diffusion are designed, based on properties of massive LOD terrain data. Using the coordinates of mid-point in a block, the index is made and divided into two parts. One part is used to index the primary patches, the other to LOD data. The index can represent the position and topological of a block very well. It is unique, and is convenient to look up. The viewpoint diffusion data retrieving algorithm use the distance between the viewpoint and the midpoint of a block to chose which block is to be rendered. Only small parts of index around viewpoint are kept in memory. So, it is not necessary to travel the whole LOD model when querying. That makes the search quick. Experiment for terrain rendering shows that, compare to traditional search in Terrain tile pyramid(TTP), using our retrieving algorithm can improve the rendering speed.(3) To support distributed storage of massive terrain data, a storage method using hybrid structure P2 P network is presented. It consists of two types of site including fundamental storage site and application site. Fundamental storage site has higher capability, and can support steady storage in P2 P network. Application site is divided into fat site and thin site, based on its performance. Such division can suit for various rendering terminal, and made the replica distribution more reasonable. A replica placement method named selective caching along road, and replica deleting method based on competitive learning are proposed. The method can guarantee distributed steady storage. It also supports replica management considering Qo S requirement and distance to fundamental sites. Experiments show that our method has better performance than traditional unstructured topology P2 P network.(4) To decrease the influence of data transmission on real-time applications, an optimized texture compression(HVS-optimized texture compression, VOTC) algorithm and an edge precision retained topography grid compression(Edge-Directed Prediction Coding, EDPC)algorithm are presented. Utilizing the property of human visual system, VOTC can guarantee less visual error of reconstructed image under the same compression ratio. For topography grid data, EDPC can achieve high accuracy of prediction and avoid the cracks caused by boundary error while decompressing data. The two compression algorithm can be implemented in parallel using GPU. They also support random access on compression data and improve the velocity of decompression. Meanwhile, the progressive transmission methods based on the two compression algorithm are designed, which decrease the waiting time of users. Experiments show that both compression methods are efficient.