基于改进局部自适应对比法的织物疵点检测  

Fabric defect detection based on improved local adaptive contrast method

作　　者：杜帅 李岳阳 王孟涛 罗海驰 蒋高明[1] DU Shuai;LI Yueyang;WANG Mengtao;LUO Haichi;JIANG Gaoming(Engineering Research Center for Knitting Technology, Ministry of Education, Jiangnan univevsity, Wuxi Jiangsu 214122, China;Key Laboratory of Advanced Process Control for Light Industry, Ministry of Education, Jiangnan University, Wuxi, Jiangsu 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214222 [2]江南大学轻工过程先进控制教育部重点实验室,江苏无锡214122

出　　处：《纺织学报》2019年第2期38-44,共7页Journal of Textile Research

基　　金：国家自然科学基金资助项目(61772238);中央高校基本科研业务费专项资金资助项目(JUSRP51727A);国家轻工技术与工程一流学科自主课题资助资助项目(2018-28)

摘　　要：为提高织物疵点检测的准确率和检测效果,采用了一种基于最相似邻域的背景估计法来进行检测。首先,利用同态滤波对图像进行预处理;然后,以滤波后图像每个像素点为中心点,以11像素×39像素的窗口大小为中心区域,通过计算中心区域与周围邻域的相似度,利用最相似的邻域进行背景估计;最后,利用背景差分原理获得目标图像,并采用阈值分割和形态学方法对图像进行处理,最终获得检测结果。实验结果表明:此方法优于传统的检测方法,不仅能够检测到复杂背景下的疵点图像,而且对不同组织及光照因素影响下的织物疵点图像同样具有很好的检测结果,检测准确率可达98%,具有较高的适用性与检出率,也具有一定的抗干扰性。In order to improve the fabric defect detection accuracy and detection effect, a background estimation method based on the most similar neighborhood patch was used to improve the detection rate. Firstly, the image was preprocessed by homomorphic filtering. Then, each pixel of the filtered image was taken as center point and window size of 11 pixel×39 pixel was taken as the central region. By calculation the similarity between the central region and the surrounding neighborhood to find out the neighborhood which was most similar to central region. So then, the purpose of background estimation was achieved. The background-difference principle was used to obtain the target image and the method of threshold segmentation and morphological was used in the image. Finally, the defection results were obtained. The experiment results show that the method is superior to the traditional detection method, not only can detection the defect image in complex background, but also has good detection results for fabric defect images under influence of external factors and different fabric weaves, the detection rate can reach 98%, and with high recognition rate, applicability and a certain degree of anti-interference.

关 键 词：织物疵点 疵点检测 自适应局部对比法 背景差分法 阈值分割 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TS181.9[自动化与计算机技术—计算机科学与技术]