A Machine Vision-based Fabric Defect Detection Solution for Textile Production Industry Using Object Detection

Along the process of textile industrial production, fabric defect detection is a critical task to ensure the reliable quality of large-scale products. To overcome the setbacks of conventional detections, there are still two crucial issues that need to be addressed. For the first time, a novel Multi-task Cascade R-CNN model based on two-stage detection method is proposed, which elegantly mitigates the large defect shape changing challenges by introducing three well-designed candidate region selection strategies and combining with cascading network to extract various defect information adaptively. As for another challenge led by fabric materials, a promoted-DSSD model on Resnet framework is proposed, which utilizes a new loss estimation and greatly refines the learning rate of unbalanced defect categories in the optimization iteration module. It overcomes the severely sample amount unbalance issue. Experimental results on the datasets collected from real textile production workshop demonstrate the excellent performance of our proposed methods, which not only meet the inspection requirements of real-time but also significantly reduce the missing rate for fabric detection in high-speed production scenarios.