基于灰度共生矩阵与反向投影的织物疵点检测

Fabric Defect Detection Based on Gray-level Co-occurrence Matrix and Back Projection

摘要:

针对织物疵点检测,将灰度共生矩阵(Gray-level Co-occurrence Matrix,GLCM)与反向投影结合起来,提出了一种基于GLCM的反向投影方法(GLCM-BP)；首先介绍了GLCM-BP的原理,然后给出了织物疵点检测流程,分析并优化了GLCM的距离d与灰度级N等参数,选择了相应的滤波与自适应阈值分割方法以检测疵点,同时给出了7种常见疵点的检测结果；最后将本文方法与GLCM方法作了检出率的比较；结果表明,提出的方法具有良好的疵点分割效果,可显著提高疵点检出率。

Abstract:

A method (GLCM-BP) which combines Gray-level Co-occurrence Matrix (GLCM) with back projection is proposed according to the fabric defect detection. Firstly the principle of GLCM-BP is introduced. Then the fabric defect detection process is given, and the parameters of distance namely d and gray levels namely N of GLCM are analyzed and optimized, and appropriate methods of filtering and adaptive threshold segmentation are adopted to detect the defects. At the same time, detection result on seven kinds of common defects is presented. Finally, this method is compared with GLCM method on the detection rate. The results show that proposed method has good defect segmentation effect and can significantly improve the defect detection rate.