Frangi滤波器和模糊C均值算法相结合的织物瑕疵检测  

Fabric defect detection based on Frangi filter and fuzzy C-means algorithm in combination

作　　者：张缓缓[1,2] 李仁忠[1] 景军锋[1] 李鹏飞[1] 赵娟[1] 

机构地区：[1]西安工程大学电子信息学院,陕西西安710048 [2]西安理工大学自动化与信息工程学院,陕西西安710048

出　　处：《纺织学报》2015年第9期120-124,共5页Journal of Textile Research

基　　金：国家自然科学基金资助项目(21301134;61301276);西安工程大学学科建设经费资助项目(107090811);中国纺织工业联合会科技指导性项目(2013066);西安工程大学大学生创新创业计划项目(201303012)

摘　　要：为解决织物瑕疵自动检测问题,提出一种基于Frangi滤波器和模糊C均值算法(FCM)相结合的织物瑕疵检测方法。首先采用均值下采样方法对采集的织物图像进行预处理,以减少织物背景纹理信息对织物瑕疵检测产生的影响;然后通过Frangi滤波器滤波增强织物的瑕疵区域;最后利用FCM处理滤波后的图像,确定织物瑕疵区域的像素和非瑕疵区域像素的聚类中心,并分割出瑕疵区域和非瑕疵区域。结果表明,本文方法检测织物瑕疵种类较多,分割效果较好。与其他方法相比,本文提出的算法利用聚类思想对织物疵点进行分割,无需利用正常织物图像进行阈值计算;另外经过滤波后疵点信息明显增强,使得疵点信息与纹理明显不同,从而使聚类更为准确,增加了检测的准确度。To solve the problem of the various defects generated in the fabric production process, a new method was proposed based on Frangi filter and fuzzy C-means algorithm （FCM） in combination to detect the fabric defects. Firstly, the fabric image sample was preproeessed using mean sampling method to eliminate the influence of background texture on fabric defects. Then Frangi filter was adopted to enhance the fabric defect area. Finally, FCM was used to determine the cluster centers of fabric defective pixels and non-defect region pixel of the Frangi filter image respectively, and defects regional and non-defective regional was segmented. Experimental results show that the proposed method could detect most of fabric defects with better segmentation results.

关 键 词：疵点检测 织物疵点 Frangi滤波器 模糊C均值聚类算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]