Comparative Analysis of YOLO Models for Real-Time Safety Helmet Detection to Enhance Construction Site Safety Using Raspberry Pi

The construction industry in India faces a high incidence of accidents and fatalities, with improper training and poor supervision being major contributing factors. The use of safety helmets is crucial in preventing head injuries, but their usage is often neglected. This paper proposes a novel approach to automate safety helmet detection using YOLOv10n, an advanced object detection algorithm, on a Raspberry Pi device. The system is designed to detect helmets in real-time, alerting workers and site managers in case of non-compliance. We compare the performance of YOLOv8n, YOLOv9t, and YOLOv10n models for safety helmet detection, and find that YOLOv10n outperforms its predecessors in terms of detection speed, making it suitable for real-time applications. Our system integrates the YOLOv10n model with a Raspberry Pi device, enabling cost-effective and efficient monitoring of construction sites. This approach has the potential to significantly reduce workplace hazards and ensure a safer environment for construction workers.