基于灰度梯度共生矩阵和SVDD的织物疵点检测  

Fabric defect detection based on gray-level gradient co-occurrence matrix and SVDD

作　　者：王孟涛 李岳阳 杜帅 蒋高明[1] 罗海驰 WANG Mengtao;LI Yueyang;DU Shuai;JIANG Gaoming;LUO Haichi(Engineering Research Center for Knitting Technology Ministry of Education,Jiangnan University,Wuxi 214122,China)

机构地区：[1]江南大学教育部针织技术工程研究中心,江苏无锡214122

出　　处：《丝绸》2018年第12期50-56,共7页Journal of Silk

基　　金：国家工信部智能制造综合标准化与新模式应用项目(工信部联装[2016]213号);江苏省产学研联合创新资金-前瞻性联合研究项目(BY2016022-35)

摘　　要：织物疵点检测是现代纺织工业产品质量控制中的关键环节之一,对保证纺织品质量具有重要的现实意义。文章基于此提出一种灰度梯度共生矩阵(GGCM)和单分类器(SVDD)结合的检测方法。该方法首先对织物原图像采用自适应中值滤波、同态滤波进行预处理,以消除图像上的光照不匀和噪声等影响,然后利用灰度梯度共生矩阵对预处理后的图像提取15个特征值并组成特征向量,经归一化后送入到单分类器SVDD中训练和测试。实验结果表明:使用此方法进行疵点检测,检验正确率达97%,漏检率为4. 5%和误检率为1. 4%,具有很好的检测效果。Fabric defect detection is one of the key links in the quality control of modem textile industry products,and has important practical significance for ensuring the quality of textiles. Based on this, a detection method combining gray-level gradient co-occurrence matrix （GGCM） and single classifier （SVDD） is proposed in this paper. In the method, firstly, adaptive median fihering and homomorphic fihering were used to preprocess the original fabric image to eliminate the impacts of illumination unevenness and noise on the image, and then GGCM was used to extract 15 eigenvalues from the images after preprocessing. The eigenvalues were then combined to form a feature vector which was normalized and sent to the single classifier SVDD for training and testing. The experimental resuhs showed that： with this method for defect detection, the test accuracy rate could reach 97% , and the missed detection rate and the false detection rate were 4.5% and 1.4% , respectively. Thus, the proposed method has a very good detection eft^ct.

关 键 词：疵点检测 SVDD GGCM 自适应中值滤波 同态滤波 

分 类 号：TS101.97[轻工技术与工程—纺织工程]