Textile defects identification based on neural networks and mutual information

Textile defects identification based on neural networks and mutual information

Abstract:

In modem textile industry, Tissue on line Automatic Inspection (T AI) is becoming an attractive alternative to human vision inspection (HVI). HVI needs a high level of human attention leading to a low level of performance in term of tissue inspection. Based on the advances in image processing and pattern recognition, T AI can potentially provide an objective and reliable evaluation on the fabric production quality. Most T AI systems claim to be able to detect the presence of defects in fabric products, and precisely locate the defects position. The motivation behind the fabric defect identification is to enable an on-line quality control of the weaving process. In this paper, a method based on texture analysis approach for identifying fabric defects in each image is proposed. Neural Networks and Mutual information are employed to characterize the texture property of a tissue image. A feature extractor is designed based on Mutual Information computation in conjunction with a classifier to minimize the error rate in defect classification.