Using Machine Learning To Classify Defects In Fabrics

Abstract

The identification and categorization of flaws is a fundamental necessity within the textile industry. Inevitable challenges arise in manufacturing processes, often resulting in the occurrence of faults or omissions in the produced goods throughout the knitting process.Therefore, it is imperative to thoroughly address textile defects via inspection activities in order to uphold high-quality standards.Fabric defect categorization refers to the systematic process of identifying underlying assumptions within diverse datasets using statistics and machine learning methodologies. The primary objective is to categorize fabrics of superior quality and devoid of defects for distribution to clients. Therefore, the 3 methods applied in this report are The Local Binary Patterns (LBP), The Gray Level Co-occurrence Matrix (GLCM) and The Local Binary Patterns (LBP) combined with a Visual Geometry Group (VGG16). The purpose of this paper is to compare and classify the effectiveness of these three approaches in determining the most optimal solution for the classify defects in fabrics. The results show that using the Local method Binary pattern LBP combined with a Visual Geometry Group (VGG16) produces the highest and most consistent fabric defect classification results. The conclusive research demonstrates that Visual Geometry Group(VGG) is an appropriate classifier for such issues. Specific faults are more easily classifiable than others. For instance, unconnected corpus, hole, and oil satin are relatively easy to categorize. In summary, the LBP technique is the most optimal approach for the identification and classification of woven faults.