Neuro-Fuzzy Fabric Defect Detection and Classification for Knitting Machine

The conventional method in textile industries for fault detection is human inspection. In this method, the classification of produced fabrics and online identification of fault in production line is done by human vision. The goal of this study is to replace the current human visual inspection methods with an automated visual inspection system that provide high accuracy of defect detection and classification. The proposed system consists of a computer based vision system (for capturing an image), image processing tools (for adjusting the image and extracting the features) and a system for detecting and classifying fabric defects. Two types of features are extracted (tonal and texture). Two types of systems are used. The first system is based on fuzzy clustering using fuzzy c-means clustering (FCM). The second system is adaptive neural-fuzzy inference system (ANFIS). Experimental results show that, the proposed systems are capable of detecting and classifying fabric defect, also it shows that using both tonal and texture features gives higher detection and classification rate than using each one individually