基于GPU加速的高分辨率实体体素化研究  

High Resolution Solid Voxelization Accelerated by GPU

作　　者：吕广宪[1] 潘懋[1] 王占刚[1] 兰向荣[1] 

机构地区：[1]北京大学地球与空间科学学院造山带与地壳演化教育部重点实验室,北京100871

出　　处：《地理与地理信息科学》2007年第1期5-9,23,共6页Geography and Geo-Information Science

摘　　要：体素化是面图形学通向体图形学的桥梁,具有广泛且重要的应用。介绍体素化方面的研究进展,分析现有体素化方法在处理高分辨率实体体素化时的不足。提出以分块为基础的高分辨率实体体素化算法:在分块内部,采用基于图形处理器(Graphics Processing Unit,GPU)的切面光栅法对分块表面体素化过程进行加速,采用射线求交方式生成种子并进行块内填充;在分块之间,设计了3方向的种子扩散面及跨块扩散机制,避免因分块而导致的大量几何求交运算,提高了效率。针对高分辨率实体体素化结果数据量大的问题,采用低分辨率表面体素化结果进行数据压缩和索引,节省了数据存储空间。测试结果表明,该文提出的算法不仅能够进行高分辨率实体体素化,而且在GIS、地学建模和CAD等体图形学相关的领域具有重要的应用价值。As a bridge between surface graphics and volume graphics,voxelization has an important and wide application in many areas. This paper summarizes the existing voxelization algorithms and their shortage in dealing with high resolution solid voxelization. Aiming at the problem in the existing voxelization algorithms,a blockbased algorithm is presented： GPU dipping method is used to voxelize the surface in every block, and ray intersection method is adopted to generate seeds filling the inner/outer space of blocks; among blocks, three seed planes are designed to help seeds span the blocks,so that a great deal computation for intersection is avoided. Based on low resolution surface voxelization, an effective method is presented to compress the large dataset of high resolution voxelization. Results show that the new solid voxelization algorithm is efficient and accurate in high resolution,and will be widely applied in GIS, geosciences modeling, CAD and other volume graphics related fields.

关 键 词：3D GIS 体素化 图形处理器 高分辨率 实体建模 体图形学 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] P208[自动化与计算机技术—计算机科学与技术]