Study Of KD-tree Based GPU Ray Tracing

Ray tracing is one of classical algorithm of global illumination in computer graphics. Itcan successfully simulate reflection, refraction and shadows and other effects. Since thealgorithm is simple in principle and easy to implement and can produce various realisticvisual effect, it has been gotten the favor of movie special effects industry, and has been usedwidely.With the rapid development of virtual reality technology and game industry, the demandof lighting effects fidelity in virtual scene is higher and higher. But ray tracing algorithm ishindered by the huge computation cost, it cannot be applied in real-time demands of virtualreality technology and gaming industry and other industries, so acceleration is the only way ofray tracing algorithm to break through the limitation. With the development of GPU parallelcomputation, GPU provides a solution strategy for a natural parallelism algorithm like raytracing algorithm.The acceleration of ray tracing algorithm is the research key in this paper, and the paperputs forward the corresponding improving algorithm, and the improving algorithm improvesthe efficiency of ray tracing. The acceleration of ray tracing algorithm is focused on the datastructure. In this paper, the different data structure is analyzed, and KD-tree structure that iscommonly used was described emphatically and the paper gets the improvement of theKD-tree. The nature of light tracing algorithm is a sample algorithm, so we put forward thecorresponding sampling strategy. Traversal algorithm is the main factor affecting theefficiency in the process of accelerated intersection; this paper summarized the work ofpredecessors, and realized the KD-tree traversal algorithm based on GPU. The paper gives thescenes simulation results of different complexity, and comparing with CPU simulation ontime, the experiment proved that the algorithm in this paper greatly improves the efficiency ofray tracing. The ray tracing algorithm based on GPU implementation in this paper can welladapt to the characteristics of the GPU architecture, and realize efficient rendering of raytracing for the size of the scene.