Efficient Real-time Detection of Complex Tire Cord Defects on Airjet Loom

The paper presents an improved YOLOv8 model tailored for the detection of defects in air-jet loom tire cord fabric, a critical component in the automotive industry. Current defect detection methods, predominantly manual, are inefficient and prone to errors, necessitating an automated solution. Leveraging advancements in machine vision and deep learning, this study presents an vision-based defect detection model that tackles the distinctive obstacles posed by the complex background, variations in target sizes, and indistinct defect features of tire cord fabric defects. The proposed model integrates a novel BiFPN (Bidirectional Feature Pyramid Network) into the YOLOv8 framework to enhance feature fusion and employs a Dynamic Head mechanism to improve scale, spatial, and task awareness. Additionally, a new loss function, Wise-Inner-Shape IoU, is introduced to refine the localization of defects. The model's effectiveness is assessed utilizing a custom dataset of air-jet loom tire cord fabric, demonstrating a notable 4.0 % enhancement over the original YOLOv8 architecture in the mean Average Precision (mAP) at a threshold of 0.5. The results show that the model is effective in achieving precise defect detection, which is essential for upholding quality standards of tire cord fabric in industrial production.