针织物疵点自动检测与分类方法的研究

Study of the Automatic Knitting Defect Detection and Its Classification

设计了针对圆筒针织物的疵点检测系统,在该检测系统的基础上,研究了一种基于小波分解的疵点检测方法和一种基于极限学习机的疵点分类方法。对采集得到的针织物疵点图像进行图像灰度转换、光照不匀校正、中值滤波等预处理,并采用Bior3.7小波分解图像,对分解后得到的纵向和横向纹理子图像提取特征值,选取针织物中常见的破洞、花针、漏针和直稀路4种疵点作为研究目标,将提取的特征值输入极限学习机中进行训练。结果表明,此方法可以有效地检测和分类白坯针织物的常见疵点。

A defect detection system for the cylinder knitted fabric was devised in this paper, on the basis of the detection system, a defect detection method based on wavelet decomposition and a defect classification method based on extreme learning machine were studied respectively. The image preprocessing such as primitive images were changed into grayscale images, the uneven illumination was corrected and images were filtered by median ill- tering, which were carried out for the knitted fabric defects image, and Bior3.7 wavelet transform was introduced to analyze the image so that the eigenvalues of image could be extracted from the transformed vertical and horizontal texture images to recognize defects, common defects like holes, pin holes, drop stitch and motion mark were cho- sen for the classification algorithm test in which the extracted eigenvalues were input into the extreme learning ma- chine for the final training. The results show that the proposed method can effectively detect and classify common defects in knitted grey cloth.