Parallel Rendering Of Large-scale DEM Based On Multi-core CPU

In recent years, researches and applications related to "Digital Earth" and "Digital City" have been given extensive attention all over the world. As the core technology, three-dimension Geographic Information System (3D GIS) also developed rapidly and become a hot research filed of "Digital Earth" and "Digital City". The topographic information is closely related with people’s lives and is a fundamental component of the three-dimensional geographic information system. Parallel rendering uses software technology to consort many graphics hardware to complete rendering tasks; parallel rendering can effectively improve the performance of rendering and meet the needs of real-time rendering of high-resolution and large-scale three-dimensional scene.As the most used client platform of3D GIS, PC platform hardly achieves smooth render large scale terrain because of the limitations of hardware and software technologies. With multi-core CPU and high-performance GPU became the mainstream of computer hardware configuration as well as the improvement of storage capacity, high-performance real-time parallel rendering on PC already has a good hardware conditions.Based on the above background, this thesis using multi-core CPU computer as implementation platform studies the large-scale DEM data scheduling algorithm, describes the terrain quad-tree LoD parallel algorithm, and discusses terrain triangulation network parallel rendering algorithm. The main content and conclusions of research are as follows:Firstly, to study large-scale DEM data organization and scheduling mechanism. Data sub-block is used as the basic unit of data block scheduling; based on the cache interval, the scheduling mechanism system provides DEM data for real-time rendering system smoothly.Secondly, to research the parallel simplify algorithm. this thesis uses quad-tree LoD to simplify terrain; terrain blocks are the basic simplify task unit and "first-come, first-served" dynamic task allocation strategy is used; different CPU core is responsible for different terrain blocks to simplify processing, to achieve parallel simplify algorithm; then to generate the topography of terrain block triangular network.Thirdly, to study the terrain triangular network parallel rendering. First obtains the number of CPU cores of the PC platform, then determines the partitioning strategy of screen according to the number; different screen area is allocated to a different CPU core and parallel render terrain triangles.At the end, to develop the large-scale three-dimensional terrain parallel rendering prototype system. Based on the multi-core CPU PC platform to achieve large-scale DEM parallel rendering prototype system and test the system performance. the memory usage of prototype system remained at about300M,frame rate stable at around30frames/sec; these indicate that the algorithm can effectively improve the performance of3D DEM rendering.In practical applications, the three-dimensional rendering of DEM is still not enough to express a realistic effect. This thesis is intended for DEM preliminary discussions on the parallel algorithm; Parallel rendering of3D terrain realistic effect also needs to address the problem of light, texture, shadows, specular reflection, and so parallel computing, but also consider how to integrate the use of vector topographic data. These are the next phase of study.