Research On Parallel Visualization Method Of Hybird Terrain Data On CPU/GPU Platform

The 3D terrain visualization technology,as one of the important research direction of the three-dimensional geographic information system(3D GIS),attracting widespread atten-tion.Traditional large-scale terrain visualization systems typically use digital elevation model which is regular GRID data structure to express three-dimensional terrain.Its struc-ture is simple but the expression data accuracy influenced by the terrain features and its expression results have load shifting "Smooth" feature.But Irregular GRID model(Triangu-lated Irregular Networks)can be retained terrain features to make up for the lack of regular GRID.Due to the huge amount of large-scale three-dimensional topographic data and the constraints of the computer hardware performance,the three-dimensional topographic data is difficult to load into memory and render in one trial.At the same time,with the develop-ment of high-performance computing hardware and software,parallel technology provides a new hardware and software technology base for a wide range terrain visualization.Based on the above background,in this paper,for a wide range of terrain,which is with high precision required three-dimensional visualization,we express the data of combining irregular GRID structure which is embedded vector feature data and regular GRID,using a multi-core CPU and a high-performance computing GPU as an achieve platform.In terms of large-scale mixed terrain data,disscussed and research on the real-time dynamic simpli-fication algorithm and the organization of parallel rendering method of the terrain,as well as the dispatch and departure method of hybird terrain.(1)GRID/TIN hybird terrain model continuous dynamic simplification algorithmResearch the simplified algorithm based on view-dependent of GRID/TIN hybrid structure terrain.In certain error threshold constraint,using viewpoint dependent algorithm to in real time simplify mixed terrain GRID,so as to improve rendering efficiency by reduc-ing the complexity of the terrain.(2)Parallel Visualization for hybird terrainResearch on parallel visualization for mixed hybird which combines parallel computa-tion and hybird terrain visualization,taking full advantage of multi-core CPU and high-per-formance GPU feature.Parallel rendering process section calculation process,thereby re-ducing the computation time consumption,and further enhance the rate of draw.(3)Dispatch and departure method for hybird terrainFor characteristics of GRID and TIN tile structure in hybird terrain,research on the data organization of methods exist inside and outside in hybird terrain,also with themulti-threaded block data prefetch buffer scheduling mechanism based multi-level buffer.According to the research,we use parallel visualization method based on hybrid terrain data.This method will block terrain data as file stored in the external memory and read data through multi-thread scheduling based on multi-level buffered.In terrain rendering,com-bined with multi-core CPU,hybird terrain adopt parallel simplified algorithm based on the ROAM and PM to real-time simplify,meanwhile use cones tailoring techniques and light algorithm based on GPU general computing to further enhance the visual efficiency,exper-iment by using embedded line features road Lushan 5 m resolution DEM mixed data.Based on verification and analysisthe of the experimental results,prove the proposed method can effectively improve the efficiency of the hybrid real-time rendering of terrain,achieve effi-cient real-time roaming in mixed terrain.