基于GPU的点云模型快速重建方法  

Fast reconstruction method of point cloud model based on GPU

作　　者：吕建新 马礼[1] 傅颖勋 李阳 马东超[1] LYU Jian-xin;MA Li;FU Ying-xun;LI Yang;MA Dong-chao(School of Information Science and Technology,North China University of Technology,Beijing 100144,China)

机构地区：[1]北方工业大学信息学院,北京100144

出　　处：《计算机工程与设计》2023年第11期3440-3446,共7页Computer Engineering and Design

基　　金：国家重点研发计划基金项目(2018YFB1800302);国家自然科学基金项目(62001007);北京市自然科学基金项目(KZ201810009011、4202020、19L2021)。

摘　　要：利用GPU并行特点,将点云数据分块并行处理,实施多线程并行重建。从有序点云的数据关系入手,设计点云处理平滑在内的数据处理方法。利用GPU对分块数据快速三角化曲面重建,使用拉普拉斯算法对重建后的曲面进行平滑处理。通过对5组不同的海量数据进行重构模型实验,与传统的单纯使用CPU重构方式相比,GPU方式将模型重构速率提升了数十倍至上百倍,模型重建速度得到了大幅度的提升,可有效适用于高精度器件装检预判,提高装配质量。The parallel characteristics of GPU were used to process the point cloud data in blocks and multi-threaded parallel reconstruction was implemented.Starting with the data relationship of ordered point cloud,a data processing method including point cloud processing and smoothing was designed.GPU was used for fast triangulation surface reconstruction of block data,and Laplace algorithm was used to smooth the reconstructed surface.The model reconstruction test was carried out through five groups of massive data with different data amounts.Compared with the traditional CPU method,GPU method improves the model reconstruction efficiency by tens to hundreds of times,greatly improves the speed of model reconstruction,and it can be effectively applied to the pre judgment of high-precision device assembly inspection,so as to improve the assembly quality.

关 键 词：有序点云 点云分块 三角化 多线程 图形处理器并行运算 表面重建 模型平滑 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]