AI-powered industrial quality assurance system for fancy yarn using computer vision and 3D visualization.

One of the problematic issues in textile production is quality control in fancy yarn because of the complicated structure and aesthetic needs of these goods. This paper presents a novel integrated system that combines artificial intelligence and computer vision algorithms for quality assurance of fancy yarns. The proposed system has been designed, implemented as a laboratory prototype, and experimentally validated under controlled conditions. The implemented prototype demonstrates capabilities for defect detection, thickness measurement, pattern analysis, and 3D visualization using a combination of 2D image processing and laser profilometry data. Under controlled laboratory conditions (22 ± 2 °C, 65 ± 5% RH), the suggested system demonstrates a defect detection accuracy of 94.7% (95%, Confidence Interval (CI) [94.1%, 95.3%]), thickness uniformity precision of 96.2%, and pattern regularity reliability of 92.5% (95% CI [91.2%, 93.8%]). The system operates on the plane yarn images in 8 states that comprise image acquisition, pre-processing, feature extraction, thickness checking, defect detection, pattern checking, quality checking and reporting. In accordance with the outcomes of the experiment, the suggested system demonstrates the accuracy of 94.7% in the defect detection, the precision of 96.2% in the thickness uniformity and the reliability of 92.5% in the pattern regularity. The integration of 3D visualization provides manufacturers with enhanced spatial understanding of yarn properties, enabling proactive quality management and reducing non-conformance to quality standards. The complete MATLAB R2024b implementation, including network architectures, training protocols, and evaluation routines, is provided in Appendix A to ensure reproducibility.