基于并行计算的抛光轴表面缺陷检测研究  

Surface defect detection of polishing shaft based on parallel computation

作　　者：姜庆胜 李研彪[1] 计时鸣[1] JIANG Qing-sheng;LI Yan-biao;JI Shi-ming(College of Mechanical Engineering,Zhejiang University of Technology,Hangzhou 310023,China)

机构地区：[1]浙江工业大学机械工程学院,浙江杭州310023

出　　处：《机电工程》2020年第12期1540-1546,共7页Journal of Mechanical & Electrical Engineering

基　　金：国家自然科学基金资助项目(51575494);浙江省自然科学基金重点资助项目(LZ14E050001)。

摘　　要：针对在生产实践中抛光轴表面缺陷检测的实时性和缺陷分类需求问题,对抛光轴表面缺陷检测系统、缺陷图像采集、图像特征、传统检测方法等方面进行了研究,归纳了轴表面缺陷检测的处理方法。提出了两种基于并行计算的检测系统,即基于云的Spark多机并行的分布式计算机集群检测系统和单机多核的CUDA并行处理检测系统;通过Spark架构多机分布式并行计算机理的分析,以及结合CUDA单机多核并行运算原理,从数学角度进行了并行化处理,最后进行了相应的实验。研究结果表明:基于CUDA的单机多核的并行处理检测系统适合实时检测,基于Spark的多机分布式并行检测系统不适合实时检测系统,但适合大数据处理系统。Aiming at the problems of real-time defect detection and defect classification requirements for polishing shaft surface defects in production practice,the surface defect detection system for polishing shafts,defect image acquisition,image features,traditional detection methods,etc.were studied,and the treatment methods of shaft surface defect detection were summarized.Two detection systems based on parallel computing were proposed,i.e.cloud-based Spark multi-machine parallel distributed computer cluster detection system and single-machine multi-core CUDA parallel processing detection system.Through the analysis of Spark architecture multi-machine distributed parallel computer theory,and analysis of the principle of mathematical parallelization combined with CUDA single-machine multi-core parallel operation theory,the experiments were conducted.The results indicate that the single-machine multi-core parallel processing detection system based on CUDA is suitable for real-time detection,and the multi-machine distributed parallel detection system based on Spark is not suitable for real-time detection systems,but is suitable for big data processing systems.

关 键 词：图像处理 SPARK 分布式并行计算 图像处理单元 CUDA 

分 类 号：TH133.2[机械工程—机械制造及自动化] TP391.4[自动化与计算机技术—计算机应用技术]