基于数字形态学滤波与SVM技术的帘子布疵点检测  

Defect Detection Based on Mathematical Morphology Processing and Support Vector Machines for Tyre Fabric

作　　者：温盛军[1] 常保磊 蒋成龙[1] 张五一[1] 

机构地区：[1]中原工学院,郑州450007

出　　处：《中原工学院学报》2016年第4期7-12,36,共7页Journal of Zhongyuan University of Technology

基　　金：国家自然科学基金项目(61304115);河南省高校科技创新团队项目(14IRTSTHN024);河南省科技攻关项目(0721002210032)

摘　　要：针对帘子布生产过程中可能出现的瑕疵问题,提出了一种多形态多尺寸的数字形态学滤波方法,采用多种不同形状和大小的结构元素对帘子布疵点图像进行滤波,对滤波后的图像使用大律法阈值分割,并提取疵点的长、宽、长宽比、面积等特征。最后,利用支持向量机(SVM)进行疵点识别。实验结果表明,该方法可准确检测帘子布中浆斑、经线粘连、断经、劈缝等主要疵点,具有分类准确、辨别速度快的优点,疵点识别率达93.3%。In view of the defect problem in production process of tyre fabric, a multi mathematical morphological structure and more size based filter method is proposed to filter the tyre fabric images with defect using different size and structure element. Then, the filtered image is separated by a big law threshold, and the characteristics of the defect are extracted, including length, width, shape features and area. Finally, support vector machine is used to identify the defect via the extracted characteristics. Experimental results show that the presented method can accurately detect the defects, such as the cord in crack, pulp spot, warp adhesion, thin warp, broken warp and weft. The defect recognition rate is 93%.

关 键 词：疵点检测 数字形态学 支持向量机 帘子布 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]