Research on intelligent detection method of safety helmet wearing on construction site based on YOLOv5

Abstract

With the increasing complexity of modern construction projects, safety risks in high-altitude operations and other special construction scenarios have become more prominent, where frequent accidents not only severely threaten workers' lives but also significantly impact construction progress and project quality. To address these safety challenges in complex environments, a safety helmet detection dataset was developed using construction site images that captures workers' wearing characteristics under various conditions, and proposed an improved YOLOv5-based detection model that specifically targets both workers and helmets while enhancing small-object detection capability for high-altitude scenarios. Through network training and performance evaluation, the model achieved 89.8% detection accuracy, demonstrating its effectiveness in enabling automated, comprehensive monitoring that significantly improves inspection efficiency while reducing safety risks and costs associated with manual high-altitude supervision. Experimental result confirm the system's strong robustness in operating reliably within complex and dynamic high-risk construction environments, meeting critical safety management requirements for specialized construction operations.

Keywords

safety helmet detection / object detection / deep learning / YOLOv5