Firefly Inspired Binary Artificial Bee Colony Based Classifier for Fabric Defect Detection

Today, the textile business is one of the most popular businesses worldwide, especially in developing countries like India, Bangladesh etc. The defect in the fabric is one of the significant factors which affect the quality of fabrics. In most of the garment industries, second quality fabrics are the result of non-detected fabric defects. The price of this second quality product is reduced by 45% – 65% of the first quality fabrics. Still, there is a need to identify the best method to inspect fabric quality efficiently, though there are quality materials and improved technologies. Most current systems use a learning-based methodology to detect flaws in basic patterned materials. In this paper, Gray Level Co-occurrence Matrix (GLCM) based co-occurrence features are extracted from the pre-processed image. And then, a novel bee colony based classifier is proposed to select an effective feature subset with the minimum number of features capable of achieving better performance than with all the features in the problem space. In specific, this research aims to improve the ability of the biologically inspired search algorithm namely Binary Artificial Bee Colony (BABC) to identify an optimal feature subset from the problem space by redirecting the individuals in the population on encountering premature convergence problems in the problem space. To offset this premature convergence problem, a novel Firefly inspired binary Artificial Bee Colony (FABC) based classifier is proposed. This classifier investigate whether Binary Artificial Bee Colony (BABC), with its exploration ability, prevents individuals in the population from getting stuck in local optima while finding an optimal feature subset from the problem space. The proposed FABC based classifier achieves 99.56% accuracy, and it is better than the existing work with an 18.18% reduction in feature size.