A Novel Fabric Defect Detection Network in textile fabrics based on DLT

Abstract

One of the most crucial and difficult computer vision tasks in textile smart manufacturing is automated quality guarantee of textile fabric materials. The algorithms for detecting flaws in fabric are examined in detail in this survey. To begin, this article provides a brief overview of the significance and inevitable arrival of fabric defect detection in the era of artificial intelligence in manufacturing. Second, there are several different types of algorithms that can be used for defect detection; these include statistical, structural, spectral, and model-based algorithms. There is an existing systematic literature review of these procedures. Finally, this research looks at how algorithms can be used to detect flaws in fabrics. Researchers and engineers in the textile industry can use this paper as a resource for learning more about detecting fabric defects. Using the average of four orientations applied to different textural features present in an image, this study proposes a method using Histogram and HF to determine the appropriate CNN with Active contour Feature for the specific type of defect. Multiple machine learning techniques were used to evaluate the constructed models. Findings from this study indicated that 85 percent of the time, participants correctly identified the relevant CNN with Active contour characteristic associated with a given textile fault in silk fabric. An accuracy of 87% is achieved by the best model generated. When we compared the same models made out of different materials, such as jute, diamond pattern fabric, and flower pattern fabric, we found that the accuracy of the diamond pattern fabric was 86% and the accuracy of the jute model was 79%.