一种基于红外成像的强反射金属表面缺陷视觉检测方法  

Vision Inspection of Metal Surface Defects Based on Infrared Imaging

作　　者：张学武[1] 丁燕琼[1] 闫萍[1] 

机构地区：[1]河海大学计算机与信息学院,江苏南京210098

出　　处：《光学学报》2011年第3期104-112,共9页Acta Optica Sinica

基　　金：国家自然科学基金(60872096);国家863计划(2007AA11Z227);中央高校基本科研业务费专项资金(2009B31914)资助课题

摘　　要：根据红外成像特性及金属表面缺陷区域灰度分布变化缓慢的特点,提出了一种基于小波纹理特性统计分析的铜带表面缺陷视觉检测方法。利用CCD视觉传感器获取受检金属表面的红外影像资料,引入一阶Haar小波分解红外图像,抽取4个小波特性,然后分别使用Hotelling T2和X2多变量统计法融合4个小波特性。最后根据统计值判别图像是否存在缺陷,并使用支持向量机对缺陷进行分类。比较分析了两种小波域多变量统计方法检测缺陷的性能。实验结果表明,Hotelling T2统计法在缺陷检测和识别方面的性能较好,对微小缺陷可达到92.8%的检测率和95.42%的识别率。According to the characteristics of infrared imaging and the gradual change of intensity levels of metal surface defects,a vision inspection method for surface defects of metal based on statistically analyzing wavelet texture has been proposed.Firstly,the CCD sensors are used to obtain infrared video-data for surface of copper strips,and then the first-order Haar wavelet is used to decompose infrared image.Secondly,two multivariate statistical methods,including Hotelling T^2 control chart and Chi square test,are used to fuse the four wavelet characteristics.Finally,the statistical values are used to distinguish the existence of defects and classify the defects using support vector machine.The capabilities of two kinds of wavelet-domain-based multivariate statistical approaches in inspecting defects have been researched deeply.The experimental results demonstrate that the Hotelling T^2 method gets the better performance,which achieves a 92.8% probability of detecting the existence of micro defects and a 95.42% probability of classifying the defects.

关 键 词：机器视觉 缺陷检测 统计分析 强反射金属 红外成像 支持向量机 

分 类 号：TP393.41[自动化与计算机技术—计算机应用技术]