Research On Ray Tracing And Parallel Acceleration Of 3D Object Model

In computer graphics,how to generate a scene with high fidelity is one of the most commonly studied problems.Ray tracing is often used to achieve high quality rendering of scene images.Ray tracing algorithm is proposed based on the principle of vision generation.It plays an important role in game industry,film manufacturing,virtual reality technology and other fields.Ray tracing algorithm is a very expensive technology,because the algorithm needs to deal with a lot of data computation.At the same time,different lighting conditions and different model structures will affect the computational cost of the algorithm.Therefore,the computation acceleration of ray tracing algorithm is also the focus of algorithm research.With the development of computer graphics hardware architecture,GPU is often used for parallel processing of various data tasks due to its powerful computing power,which has obvious advantages over traditional CPU methods in computing performance.This provides a new direction for the research of ray tracing algorithm acceleration methods.In this paper,the main object is the acceleration method of ray tracing algorithm,which is based on the traditional ray tracing algorithm.Scene management is the most commonly used method to improve the efficiency of ray tracing algorithms.Different scene management methods use different rules to divide the space scene,and all of them reduce unnecessary intersection calculation by screening and tracking rays to accelerate.Kd-tree is the most efficient partition method for ray tracing algorithm in static scene.This paper first introduces the basic principle and rendering process of ray tracing.With focusing on the scene management,the KD-tree structure is constructed by CUDA programming platform.Different models are used to verify the acceleration effect of KD-Tree structure.Then,a combination of scene management method is suggested to solve the influence of traditional segmentation methods on the structure stability of KD-tree.The method of combining LBVH and KD-Tree not only ensures the efficiency of scene space partitioning but also meets the requirements of KD-tree structure stability.Finally,the experimental results show that the combined scene management method is more stable than the GPU ray tracing acceleration algorithm under KD-tree structure partition.