The Mechanism And Realization Of Parallel Rendering Of OpenProbe Seismic Volume Data

With the development of Three-Dimensional graphics technology and thevisualization technology, the size of data increases exponentially. Although the singleCPU processing power is growing rapidly, but compared with the size of the data’sincreasement, it’s very slow. It’s also very normal that a single machine render a largedata with difficulties. At the same time, the professional-level image processingservers and workstations can meet the demand that users require, but it cannot bewidely used due to the high price.The3D geological visualization technology has become more and more popularin the real usage of geoscience. On one hand,the visualization of geological data canprovide true information of the deep earth to the scientific research faculties, toprovide reference of earth structure. On the other hand, it can also help the detectionof energy resources, producing economic benefits. The geological data generated inthe process of exploration is so large that the size of it can be reached to dozens of GBor even TB. But in the process of data visualization, the data can hardly be whollyloaded to the memory of PCs, which leads to the display process deficiencies andlow-speed processing.Scene graph, the most important method in Three-dimensional scenemanagement, is widely used nowadays. It can be easily constructed using thehierarchical structure of each complex scene, and scene graph is based on OpenGL.So, the parallel rendering based on high-speed networks cluster is quite cost-effectiveand extensible.This article is based on the optimization of major national special fund "DeepExploration of Earth" which developed the Three-dimensional earth data visualizationtoolkit--OpenProbe. OpenProbe is a package used to render image for large-scale data on a single machine. The paper mainly builds a cluster, and through multipleprocesses and MPI libraries, achieving parallel rendering based on scene graph.The main work of this paper is divided into following aspects: firstly introducingseveral concepts about parallel rendering and some parallel rendering systems basedon scene graph, analyzing the advantages and disadvantages. Secondly，introducingthe fundamental state of OpenProbe, this three dimensional geoscience datavisualization toolkit, showing the architecture and describes the functions and analyzethe optimizing parts of OpenProbe, focusing on multi-thread data loading and parallelrendering process which can be optimized, realizing the optimization of data loadingprocess and the function of parallel rendering.Finally, this paper mainly uses different sizes of data as the testing data,comparing the costing time of the same data using different methods, I set multipletests. The experimental results show that, in a distributed cluster environment, thecosting time for rendering has lowered a certain extent.