A SURVEY ON WOVEN FABRIC DEFECTS

• Main Authors: YAŞAR ÇIKLAÇANDIR Fatma Günseli, UTKU Semih, ÖZDEMİR Hakan
• Format: Article
• Language: English
• Published: Editura Universităţii din Oradea 2019-05-01
• Series: Annals of the University of Oradea: Fascicle of Textiles, Leatherwork
• Subjects: Image processing; textile; woven fabric; clustering; classification; fabric defects
• Online Access: http://textile.webhost.uoradea.ro/Annals/Vol%20XX%20nr.%202-2019/Textile/Art.%20no.%20388%20pag%20113-118.pdf
• ISSN: 1843-813X; 2457-4880

There are many studies in the literature to find and identify woven fabric defects. However, the systems developed have been designed to identify only certain defects due to the large number of defect types. Moreover, a method that works well for identifying a defect type may not work for another defect type. In addition, some mistakes are easy to recognize, while others are difficult to recognize. In this study, different clustering and classification techniques have been investigated and studies detecting and recognizing fabric defects using these techniques have been investigated. Clustering algorithms are often used to detect defects while classification methods are used to recognize the defect types. When we look at the literature, the most common clustering algorithm for fabric defect recognition is K-means algorithm, while the most common classification technique is neural networks. In general, neural networks have been used in the vast majority of studies. The automatic recognition of fabric defects has not yet achieved the desired level of success. Approximately 80 percent of studies conducted on this field have only developed a model, but have not compared the method they used with other methods. So, very little work has been tested in more than one method.