Detection of Defects on Warp-knit Fabric Surfaces Using Self Organizing Map

Abstract

Warp-knit fabric surface is a patterned textured surface with repetitive units on its surface. Only few researches have been reported for detection of defects on patterned fabric surfaces. In this paper, an anomaly detection method based on self organizing map is proposed for detecting defects on warp-knit surfaces. The method consists of self organizing maps on two levels. The method was applied to a set of images of 8 different types of warp-knit surfaces, which included samples from the 8 categories of defects. According to the experimental results, defect detection rates of proposed method are close to 80 percent in most categories of defects.