Enhancing YOLOv8 with Global Attention Mechanism for Bolt-Missing Detection in Industrial Environments

To address challenges in industrial Bolt-Missing detection-including difficulties in identifying micro-objects, sensitivity to environmental disturbances, and insufficient real-time performance—this study proposes an enhanced YOLOv8-based intelligent detection method. By integrating a Global Attention Mechanism (GAM), Transfer learning, and multisource data augmentation, the method significantly improves detection robustness in complex environments. The GAM module enhances feature representation by capturing channel-wise and spatial interactions critical for bolt thread recognition. Experimental results demonstrate state-of-the-art performance with 88.23% recall, 92.39% [email protected], and 68.57% [email protected]:0.95, outperforming conventional models. Subsequent research will prioritize lightweight model optimization to achieve real-time detection capabilities for both assisted inspections and autonomous monitoring in industrial environments, thereby enabling scalable real-time surveillance solutions.