Semantic Web-Based Image Super-Resolution for Enhanced Steel Surface Defect Detection

This study addresses the challenges of steel surface defect detection by introducing the ATG-SRGAN model, which integrates Semantic Web technologies with an Adaptive Transformer and Generation module. By leveraging self-attention mechanisms and residual convolution within the SRGAN framework, the model enhances image resolution, reduces noise, and uncovers finer details, achieving superior PSNR (37.17) and SSIM (0.9590) metrics. Integrated with an improved YOLOv8-ghost-p2 framework, the model also attains a mean Average Precision (mAP) of 81.2 and a recall rate of 76 in defect detection. These results demonstrate the model's potential to significantly improve automated quality control in industrial settings, highlighting the importance of Semantic Web technologies in advancing defect detection methods.