Study On Computer Vision-Based Automatic Fabric Grading

In the whole process of fabric manufacturing, fabric inspection is an important procedure. So far, the traditional fabric inspection is completed manually, which is a dull, tedious and tiring work. The result is easily influenced by the inspector's ability, his mentality condition and his physiological state, meanwhile, it is inefficient. With the development of computer and digital image processing technique, the computer vision is more and more frequently applied in industrial production. The computer vision-based automatic fabric inspection has become one of the research focuses.Based on wavelet transform-based automatic fabric defects detection, namely defects have been detected, with image processing technique, the automatic fabric grading is discussed, the specific contents are as follows.1. With image processing technique, the defect segmentation and characterizing are discussed. Firstly, the regular texture on fabric surface is smoothed with medial filter, secondly, the defect edge is inspected with Sobel operator, and then, by magnifying the image gray histogram, the threshold is found to segment defects, finally, the exact defect binary images are gotten using morphology processing to remove noise points around defect. By computing defect binary images to extract feature parameters, defects are characterized.2. The rules of automatic defect assessment and fabric grading are made. Based on the characteristic of wavelet transform-based automaticdefects detection and of defects segmentation with image processing technique, the rules of automatic fabric inspection are made according manual rules.3. Based on wavelet transform-based automatic defect detection and defect segmentation with image processing technique, the software of automatic fabric grading is established primarily, and carried out with Microsoft Access 2000 and Microsoft Visual Basic 6.0.We got conclusions as follows.1. The automatic fabric grading system is abstractly feasible, which is composed of three parts: defects detection, defects segmentation and characterizing, defects assessment and fabric grading. It is also proved correct and practical in practice.2. The automatic segmentation, characterizing and assessment results of some typical defects yielding during routine weaving process are reported, such as coarse end, missing end, double warp, double pick, taped filling, open setting, slub yarn, gout, fabric hole, oiled defect, weft kinks, tight thread and etc. The experimental results show that these typical defects can be segmented and characterized correctly with image processing technique discussed in this paper.3. For those defects, which are identified as the same type in manual and automatic rules, the automatic defect assessment rules are the same as manual, for those defects, which are identified as the different type, new rules are made according to manual. The defects, which are identified as the different type, are assessed with automatic rules and manual rules respectively, the experimental results show that, with automatic rules made in this paper, defects can be assessed in accord with manual rules basically.4. Using the automatic fabric grading software, the defects are segmented and assessed, the fabric is graded. The experimental resultsshow that, the technique of automatic fabric grading discussed in this paper is practical.