An Approach to Fabric Defect Detection using Statistical Methods for Feature Extraction

Abstract

Quality control in any commodity at every stage of production not only is desired but its reproduction day in and day out is essential in this fast changing world especially when we see that shopping from a digital screen is the new trend in this era of pandemic. Quality check, control and inspection demands manual checking where in the monotonous work can make a human negligible towards work and lead to reduction in quality control. The world demands replacement of such laborious work which can only be promised by machinery custom tailored from the needs of providing accuracy, time reduction in quality checks as well as production and most importantly reduction in labour cost when commodity manufacturing is concerned. One such commodity that we cover in this paper is the textile industry where in automated fabric defect detection is harnessed as the research trend across the world with the help of image processing combined with machine learning algorithms is concerned. This area of research is challenging as the algorithm needs to cover robustness, efficiency and particularly the variety of classifiable defects in terms of surety and complexity. Out of the various techniques utilised for fabric defect detection we have utilised a threshold based image processing technique that can provide an automated stepping stone for atomization in the fabric defect detection for quality control.