基于最优参数非线性GLCM的织物瑕疵检测算法  

Fabric defect detection algorithm based on nonlinear GLCM with optimal parameters

作　　者：董蓉[1] 李勃[2] 

机构地区：[1]南通大学电子信息学院,江苏南通226019 [2]南京大学电子科学与工程学院,江苏南京210093

出　　处：《计算机工程与设计》2015年第9期2467-2471,2507,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61401239);南通市科技基金项目(BK2014066)

摘　　要：传统织物瑕疵检测多为人工操作,存在主观性强、效率低、成本高等缺点,为此提出一种基于最优参数非线性灰度共生矩阵(GLCM)的织物瑕疵自动检测算法。将传统的线性GLCM构建方式改为非线性,使构建的GLCM能更有效捕捉图像特征、区分瑕疵;通过对无瑕疵织物图像的学习,获得非线性GLCM特征提取的最优尺度方向参数以及自适应的瑕疵分割阈值;采用获得的参数提取待检测图像的特征,通过特征相似性距离度量定位瑕疵区域。针对实际图像的实验结果表明,该算法能有效定位织物瑕疵区域且受噪声干扰小。Traditional fabric defects detection is often operated manually, therefore shortcomings exist such as strong subjectivity, low efficiency and high cost. To solve these problems, an automatic fabric defect detection algorithm based on nonlinear gray level co-occurrence matrix （GLCM） with optimal parameters was proposed. Traditional linear GLCM construction method was modified to a nonlinear way so that the established GLCM captured image features and distinguished defected area more effective- ly and efficiently. The optimal scale and orientation parameters for nonlinear GLCM feature extraction and an adaptive threshold for defected area segmentation were obtained by training undefected fabric images. Those parameters were used to extract features from images to be detected and defected areas were located by feature similarity distance measurement. Experimental results for real images show that the proposed algorithm can effectively locate fabric defect areas and be insensitive to noise.

关 键 词：织物瑕疵检测 特征提取 灰度共生矩阵 相似性度量 最优参数 

分 类 号：TN911.73[电子电信—通信与信息系统]