Development of an intelligent fabric defect inspection system

The aim of this study is to design a prototype machine vision system for the fabric inspection machines and develop algorithm for fabric defect detection and classification automatically. The material is selected as `undyed raw denim? fabric. Thus, the fabric inspection process could be achieved in shorter time with higher performances and evaluation of the fabric defects could be performed objectively. The system could be easily adapted on the existing fabric inspection machines. Three different defect detection algorithms were used; Linear Filtering (LF), Gabor Filter (GF) and Wavelet Analysis (WA). These algorithms were applied off-line and real-time over five types of defects; warp lacking, weft lacking, hole, soiled yarn and yarn flow or knot. Defect database was prepared for off-line applications by scanning the fabric samples. Experimental set-up was built for real-time application. User interface was prepared for each real-time algorithm application. The performance and the success rate of the applications were evaluated and discussed. The defective fabric images were then classified by using Artificial Neural Network (ANN) method. A user interface was also formed for this application. The classification rate of the algorithm was evaluated statistically at the end.