首页|基于GPU的实景三维模型裁剪算法研究

基于GPU的实景三维模型裁剪算法研究

扫码查看
图形处理器(Graphic Processing Unit,GPU)作为主流高性能计算的加速设备,已越来越多地应用于诸多领域的并行计算中,利用GPU的并行计算能力,可以极大地提高传统算法的计算效率。文章主要研究GPU多线程计算方法与统一计算架构(Compute Unified Device Architecture,CUDA)技术在实景三维模型裁剪中的应用,提出了一种基于GPU的实景三维模型裁剪算法,包括设计了基于面拓扑的多级索引结构,以实现线程内重复交点快速查找;提出了一种轻量多边形三角化方法,优化算法流程;使用多种优化策略,在不影响裁剪网格质量的情况下进一步提高算法的性能。结果表明:根据模型大小与裁剪次数的不同,相较于传统算法,所提方法在单次裁剪的情况下加速比可达 13。93,在多次裁剪的情况下加速比可达 35。85,显著地提高了模型的裁剪效率。
Research on real 3D model clipping algorithm based on GPU
As a mainstream high-performance computing acceleration device,Graphic Processing Unit(GPU)has been increasingly used in many fields of parallel computing,and the computational efficiency of traditional algorithms can be greatly improved by using the parallel computing capability of GPU.The paper focuses on the application of GPU multithreaded computing method and Compute Unified Device Architecture(CUDA)technology in the real 3D model clipping,and proposes a real 3D model clipping algorithm based on GPU.The main work includes:Designing a multi-level index structure based on face topology to realize the fast search of repeated intersections in threads;Proposing a lightweight polygon triangulation method to optimize the algorithm flow;Adopting varying optimization strategies to further improve the performance of the algorithm without affecting the quality of the trimmed mesh.The results show that the acceleration ration of the proposed GPU-based method can reach 13.93 under single clipping and 35.85 under multiple clipping according to the difference of model size and clipping times which significantly improves the clipping efficiency of the model.

graphic processing unitreal 3D modeltriangulation clippingparallel computing

马东岭、李铭通、朱悦凯

展开 >

山东建筑大学 测绘地理信息学院,山东 济南 250101

图形处理器 实景三维模型 三角网裁剪 并行计算

山东省自然科学基金项目山东建筑大学博士科研基金项目

ZR2020MD025X21079Z

2024

山东建筑大学学报
山东建筑大学

山东建筑大学学报

影响因子:0.576
ISSN:1673-7644
年,卷(期):2024.39(1)
  • 8