Abstract:

When the image processing method is used to detect the yarn evenness, the background blackboard as well as the yarn hairiness and the image noise would have great image noise would have great influence on the detection results. To solve this problem, a method referring to the human visual perception mechanism for detecting yarn evenness based on saliency algorithm was proposed. Firstly, the color and brightness features were extracted from the collected yarn image saliency of for saliency analysis to highlight the yarn evenness area. Then the iterative threshold segmentation algorithm and the area filtering were adopted to obtain accurate and clear yarn evenness binary images. Based on the binary images, the diameter and yarn evenness were calculated, and the yarn defect was determined. The edge accuracy evaluation shows that the proposed method of saliency analysis can obtain the yarn evenness binary images with better segmentation. Compared with the evenness detection result of the Uster Classimat 5, the results obtained by the method are accurate and have a good consistency with those of the Uster Classimat 5.

Key words: yarn evenness, saliency analysis, iterative threshold segmentation, area filtering