Abstract
Ensuring compliance with personal protective equipment (PPE) requirements remains one of the most persistent challenges in construction site safety management, and conventional manual inspection struggles to provide continuous, site-wide coverage. This study evaluates the field performance of an unmanned aerial vehicle (UAV)-based monitoring system that combines drone imagery with computer vision to automatically verify PPE compliance among construction workers. The object of the study is construction-related PPE, and the subject of the study is the effectiveness of its detection by computer vision using a UAV camera. The system developed by the authors was tested using a DJI Matrice 4E UAV (Da-Jiang Innovations Science and Technology Co., Ltd., Shenzhen, China) at an active construction site. Based on the resulting video recordings, a dataset of more than 500 high-resolution images was compiled and manually annotated to support training and evaluation. Field performance was assessed across four scenarios monitoring four to seven PPE items simultaneously, using the percentage of matches against manually annotated reference images, descriptive statistics, and trend analysis, and the results were discussed against comparable field and benchmark studies. The average number of correctly detected objects ranged from 3.75 to 5.75 across the four scenarios, corresponding to a relative detection rate of 82.1% to 93.8% of each scenario's maximum, with a coefficient of variation not exceeding 17.8%, supporting the system's potential as a continuous, site-wide complement to manual safety inspection.