基于CUDA的光线追踪优化算法研究与实现

ON CUDA-BASED OPTIMISED RAY TRACING ALGORITHM AND ITS REALISATION

在三维场景仿真过程中,为了实现真实的光影效果,通常采用光线追踪法对场景进行渲染。光线追踪算法的核心过程是光线与场景中的片元进行相交测试,而对于一个复杂的场景,该过程计算量非常大。为了改善光线追踪算法的计算速度问题,实现一种基于CUDA(Compute Unified Device Architecture)的光线追踪算法。该算法利用GPU的并行处理能力同时结合KD-Tree加速相交测试过程,最终提高仿真场景的渲染速度。通过实验表明,该算法的KD-Tree创建性能相比传统方法提升约20%,光线追踪性能提升约6倍。

In three-dimensional scene simulation process, in order to achieve realistic lighting effects, ray tracing method is usually used to render the scene. The intersection test of the light and the render-element is the core process of ray tracing algorithm, and for a complex scene, the process is highly computational. We implement a ray tracing algorithm which is based on CUDA （Compute Unified Device Architecture）. The algorithm makes use of GPU＇s parallel processing capability, and combines with KD-Tree to accelerate the intersection testing process at the same time. It is demonstrated through experiment that our algorithm improves the KD-Tree creation performance about 20% compared with traditional method, and the ray tracing performance is also improved about 6 times.