Abstract: In this paper, an approach based on digital image processing techniques to detect irregularity of yarn was proposed. Four different specifications of pure cotton yarn are taken as an example. Firstly, the yarn images were captured with MRS-4800M48U scanner. Then, the yarn images were treated sequentially with Wiener filtering, binaryzation and morphology opening operation. The clear images were obtained with no noise and the yarn diameters and the CV values of yarns can be calculated according to the treated images. The CV values based on this image approach were approximated to the results tested by Uster Tester. The experimental results show that the primary approach proposed in this paper to detect yarn evenness is feasible; the CV values of yarns can be detected by this digital method accurately and efficiently, which can show the yarn appearance quality objectively.

Key words: yarn evenness, diameter deviation, Wiener filtering, threshold segmentation, morphological operation