An Efficient Real-Time Safety Helmet Detection Algorithm Based on YOLOv11

To address personnel safety management challenges in large-scale engineering scenarios such as mines, rail transit, construction sites, and fossil power extraction, this study proposes a safety helmet detection algorithm. Built upon the YOLOv11s model, the algorithm incorporates the EMA attention mechanism and refines C3k2 to C3k2-EMA, thereby improving the model's feature extraction capabilities. Furthermore, structured pruning is applied to reduce the number of parameters and floating-point operations, which significantly enhances the model's running speed without compromising accuracy. Experimental results demonstrate that the proposed algorithm can reliably detect whether personnel at construction sites are wearing safety helmets, satisfying real-time requirements and making it suitable for deployment in engineering monitoring equipment.