A fuzzy system for detection and classification of textile defects to ensure the quality of fabric production

Abstract

The aim of this research focuses on construct a computerized system for textile defects detection. The system merges between image processing methods, statistical methods in addition to the Intelligent techniques via Neural Network and Fuzzy Logic. Gabor filters were used to identify edges and to highlight defective areas in fabric images, then to train the neural network on statistical and geometry features derived from fabric images to form the special neural network distinguish and classify defects into the fourteen categories, which are the most common defects in the textile factory.  The proposed work includes two phases. The first phase is to detect the defects in fabrics. The second phase is the classification phase of the defect. At the defect detection stage, a Discrete Cosine Transfer (DCT) converts the images to the frequency domain.  Image features then drawn and introduce them to the Elman Neural Network to detect the existence of defects. In the classification stage, the images are converted to the frequency domain by the Gabor filter and then the image features are extracted and inserted into the back propagation network to classify the fabric defects in those images. Fuzzy logic is then applied to neural network outputs and interference values are used in fuzzy logic to increase final discrimination. We evaluate a distinction rate of 91.4286% .After applying the fuzzy logic to neural network output; the discrimination rate was raised to 97.1428%.