Enhancing Construction Safety Monitoring Through Yolo-Based Ppe Detection Systems

Abstract

Non-compliance with the use of personal protective equipment (PPE) remains a persistent safety issue during construction project implementation, contributing to a high incidence of site accidents. While prior studies have extensively highlighted the importance of PPE use, enforcement, and monitoring practices on construction sites, these practices continue to rely heavily on manual supervision, which is inconsistent, labour-intensive, and prone to oversight. A critical gap, therefore, exists in the continuous, objective monitoring of individual workers' PPE compliance, particularly in complex, dynamic site environments. This study proposes an artificial intelligence–based PPE monitoring system using the You Only Look Once (YOLO) object detection algorithm to automatically identify the use of safety helmets, safety footwear, and reflective vests on construction sites in Malaysia. The research focuses on evaluating the detection model's performance in terms of accuracy and its ability to provide timely information to personnel responsible for safety supervision. Model performance was assessed through detection accuracy and system responsiveness during controlled testing. The findings indicate that the proposed model can identify PPE non-compliance with a satisfactory level of accuracy, demonstrating its potential as a supplementary tool for site safety monitoring. Although the system does not replace human supervision, it provides a foundation for automated safety inspection and supports proactive safety management. The study contributes to ongoing research on digital safety monitoring and highlights the role of computer vision in strengthening construction site safety practices.