Research On Distributed Real-time Rendering Technology Of Super Large 3D Scene Model

In recent years,virtual simulation,aerospace scientific exploration,scientific computing visualization and many other fields tend to be refined in 3D scene display,so the model construction is more and more complex,the model volume is also gradually large,including vertices and triangles to reach tens of millions or even hundreds of millions.In the face of such a large volume of model,ordinary single PC or single graphics card cannot bear such a large amount of computation,in this case,it is inevitable that the 3D scene model cannot be rendered,which makes the research in the above fields difficult to progress.Multi-machine collaborative distributed real-time rendering can take advantage of multi-processors and multi-graphics,increase the overall rendering rate,and meet the computing task requirements of large-scale models.This paper conducts related research on distributed real-time rendering of super-large 3D scene models,aiming to explore new rendering strategy.The purpose is to provide a new idea for calculation and rendering of super large 3D scene model.The main work and innovations of this paper are as follows:(1)Comparing the performance of octree and k-d(k-dimensional)tree when segmenting,loading and rendering large-scale models,it is concluded that k-d tree is a better data structure for real-time rendering system through analysis.Introduce the specification of the 3D point cloud ply(polygon)file format,design the k-d tree partition and storage algorithm of the data in the file,and provide a research and construction plan on the data structure for the establishment of the rendering system.(2)Analyze the characteristics of the traditional master-slave architecture,and design three types of system role architecture construction plans on this basis.The internal nodes of the system are divided into task computing end,master control rendering end and slave rendering end.The structure is optimized The workload of the master node in the master-slave system structure is balanced,and the internal node tasks of the system tend to be balanced,which provides a new idea for the construction of a distributed real-time rendering system.(3)According to the frustum expansion technology in the field of visual simulation,the task division strategy based on frustum segmentation and the kd tree retrieval strategy are designed to realize the task division of the scene model and apply it to the task assignment of the distributed real-time rendering system,Has proved its effectiveness through experiments.(4)The Five-hundred-meter Aperture Spherical Radio Telescope(FAST)three-dimensional scene model with the number of vertices and triangles exceeding10 million levels is used for multi-screen display and single-screen display verification.The experiment will follow the rendering The number of nodes increases,rendering efficiency improves faster,and the system design goal is achieved.