基于混合数据结构的实时GPU光线跟踪算法  

Coupled Use of BSP and BVH Trees in GPU-based Real Time Ray Tracing

作　　者：谢智歌[1] 熊岳山[1] 陈长波[1] 王彦臻[1] 

机构地区：[1]国防科技大学计算机学院,湖南长沙410073

出　　处：《计算机仿真》2010年第12期232-236,共5页Computer Simulation

基　　金：国家自然科学基金项目(60773022);国家863计划项目(2007AA01Z313)

摘　　要：研究真实感图像的问题,光线跟踪算法是生成真实感图形的主要算法之一。为了解决光线跟踪算法计算量大、效率低下、对动态场景处理能力不足的问题,采用图形处理器(graphics processing unit,GPU)统一编程架构,提出了一种新的基于计算统一设备架构(Compute Unified Device Architecture,CUDA)编程模型的结合层次包围体(Bounding Volume Hierarchies,BVH)和kd-tree的光线跟踪方法,首先对场景进行启发式的层次包围体分割,随后在每一个层次包围体上做启发式的k维二叉树(k-dimensional tree,kd-tree)剖分,这样就形成了一种新的空间划分,有利于建立高效的空间数据结构,提高光线与场景物体的求交速度。新设计的算法易于扩展到动态场景的光线跟踪,而且具有更快的渲染速度,有利于在应用系统中集成和扩展。实验结果验证了算法的有效性。One of the most popular and useful ways of generating realistic images is ray tracing.In this paper,we adopt unified programming architecture on GPU,and then present a novel ray tracing algorithm that coupled use Bounding Volume Hierarchies（BVH） and kd-tree based on CUDA programming model.In our algorithm,we first divide the scene by heuristic BVH,then subdivide every bounding volume with the method of heuristic kd-tree.Finally,we obtain a new spatial subdivision,which is beneficial to build efficient spatial data structure and speeds up intersection testing between ray and scene.Our new algorithm can easily extend to dynamic ray tracing,and smoothly be integrated with application system.As a result of above all,we obtain higher rendering speed.The experimental results show that the novel approach is effective.

关 键 词：光线跟踪 二叉树 层次包围体 图形处理器 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]