Development of Helmet Detection and Classification Technology to Enhance the Safety

Abstract

This research presents the development of an artificial intelligence model for object detection, specifically for identifying and classifying Personal Protective Equipment (PPE), focusing particularly on detecting safety helmets in construction site contexts, which have high risks and require strict safety measures. The main target group is construction workers, who are most prone to accidents. The developed model can detect the wearing of safety helmets in both general cases and instances where workers wear sun hats underneath safety helmets, which is common in Thailand's context. Furthermore, the model can accurately classify the colors of safety helmets to identify the wearer's role, such as yellow for construction workers and white for engineers. The development utilizes the YOLOv9 (You Only Look Once), which is one of the highly efficient deep learning techniques for object detection. The construction worker photographs were collected covering various camera angles and real situations to create model weights that can be effectively applied in automated safety inspection systems at construction sites. The model evaluation results show high precision (90.5%) and good mean Average Precision (mAP@50 84.8%) values, indicating the model's potential to appropriately detect and classify safety helmets for construction workers in complex environments.