| In the process of the production and checkout of textile and garment industry, the textile fabric surface defects inspection and recognition have been conducted for a long time at home and abroad. But it still relies on manual checking in textile garment processing industries. As we all know, the defects are the main factors which affect the quality of garments. In order to guarantee the quality of garments, each garment must be checked before leaving factory, so garment defects checkup is one of the most important links of garment quality control and management. For a long time, garment defects checkup are detected by artificial vision, but there exist the shortcoming of labor-intensive, low-speed detection, liable to be affected by human factors, high rate of missing and leaking defects, inconvenience of data processing to use traditional detection methods. So it needs to develop novel, rapid and accurate automatic clothing defects detection system.Applying image processing and pattern recognition technology to the checkup of garment defects are less researched at home and abroad, in particular on garment sewing technics. In order to realize automatic detection of garment defects, the paper does some exploring research to garment seam defects recognition and classification using digital image processing and pattern recognition technology. The paper includes system architecture analysis of garment seam defects system, garment seam samples collection and classification, the pretreatment and segmentation of garment seam defects, the characters extraction of garment seam defects, the research of garment seam defects recognition and classification.The paper does research on four types of garment seam defects of image processing and the algorithms of pattern recognition, and it does the following work:Firstly: Image collection and classificationIn the paper, garment seams and the types of defects are defined and classified in detail. The images of garment seam sample are normalized, and the paper designs the sample image templates of simple chain seam and dual-chain seam separately. The task collects four types of garment seam in total, and collects the common types of defects aiming at specific types of garment seam.Secondly: Putting forward the corresponding image processing method according to the different types of seam In the paper, the image processing includes image preprocessing and image segmentation. It puts forward corresponding method of image preprocessing and image segmentation which aims at different types of garment seam defects and apparel fabrics through a number of emulational experiment, and it reserves the useful garment seam information more completely, wipes off the unwanted fabric texture and other noisy information. The chapter has done a good job fully prepared for the following feature extraction.Thirdly: Feature extraction of garment seam defectsThe chapter extracts the effective characteristic parameters according to different types of garment seam and the types of its defects, and then it does more scientific processing. The chapter does preparation job for the latter recognition and classification.Fourthly: The research of garment seam defects recognitionThe chapter researches the reasonable defects recognition algorithms which aim at the different types of seam defects according to the characteristic parameters and pattern recognition methods, and it achieves the goal of recognizing different types of seam defects.The main contribution of the paper:(1) It does detailed classification of garment seams in collecting samples stage, and it designs the templates of seam sample images which make all garment seam samples normalized. (2) It researches the corresponding image processing method according to four types of garment seam samples, and it realizes the purpose of separating garment seam and background texture clearly.(3) It extracts the characteristic parameter which can reflect the feature of seam defects according to the character of garment seam, and it does effective disposal to these characteristic parameter.(4) It researches the corresponding algorithms of recognition and classification respectively which aim at four types of garment seam samples through a number of exploring experiments, and it does the detailed analysis and explanations to these recognition and classification algorithms. |
