基于GPU的金属工件表面缺陷检测算法的研究  

RESEARCH ON METAL WORKPIECE SURFACE DEFECTS DETECTION ALGORITHM BASED ON GPU

作　　者：蒋财运 刘羽[1,2] 韦树烽 

机构地区：[1]桂林理工大学广西矿冶与环境科学实验中心,广西桂林541004 [2]桂林理工大学机械与控制工程学院,广西桂林541004 [3]桂林理工大学信息科学与工程学院,广西桂林541004

出　　处：《计算机应用与软件》2015年第3期271-274,共4页Computer Applications and Software

基　　金：广西教育厅科研项目(教育201102ZD018);2012年研究生教育创新计划项目(YCSZ2012085)

摘　　要：在金属工件的生产过程中,不可避免地会生产出一些不良品,必须进行快速识别。缺陷检测系统需使用图像采集设备采集金属工件的表面图像,完成混合噪声滤除等预处理后,进行图像配准并使用差影法分割图像,然后标记缺陷和提取缺陷纹理特征,最后进行工件缺陷的分类和识别。为了提高金属工件表面检测系统的检测速度,以满足高速生产流水线对检测系统的高实时性要求,依托GPU平台设计了一套合理的并行算法来完成不合格工件的自动检出工作。实验结果表明,在满足检测精度的前提下,基于GPU的并行图像处理算法相对于串行算法能取得较好的加速效果（实验环境为3.2~10.3倍加速比）,为工件表面缺陷的快速检测提供了一种新的途径。It will inevitably produce a number of defective goods in the process of metal workpiece production,and they have to be rapidly recognised. The defect detection system need to use image acquisition device to capture the surface images of metal workpiece first,after the pre-treatment of mixed noises filtration is done,the image registration is carried out and the difference image method is employed to segment the image; then the defects are marked,and the texture features of the defect are extracted; finally the classification and recognition of workpiece defects are processed. In order to improve the detection rate of metal surface detection system so as to meet the high real-time requirement of high speed production line on detection system,we design a set of reasonable parallel algorithms based on GPU platform to do the automatic detection work for unqualified workpieces. Experimental results show that on the premises of satisfying the detection precision,the GPU-based parallel image processing algorithm achieves better speedup effect relative to the serial algorithm（ 3. 2 ~ 10. 3 times in our experimental environment）,which provides a new approach for fast detection of workpiece surface defects.

关 键 词：缺陷检测 数字图像处理 并行计算 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TH161.14[自动化与计算机技术—计算机科学与技术]