Intelligent safety monitoring system at construction sites

This article discusses current approaches to the use of computer vision technologies in the tasks of ensuring safety at construction sites. Special attention is paid to methods of detecting personal protective equipment, neural network architectures used for visual data analysis, as well as practical aspects of integrating such solutions into real production processes. An automated control system based on the YOLOv12 artificial intelligence model architecture has been developed. The system is designed to analyze the video stream in real time in order to identify violations related to the lack of personal protective equipment, in particular construction helmets. There are three main modules implemented in the system. Video data processing modules, detection and classification of violations, and automatic report generation. The training of the model is based on real video data, which ensures adaptation to the conditions of the construction site with variable lighting, overlapping objects and background noise. The limitations of the algorithms are related to low visibility areas and partial overlap of objects. In the future, it is planned to expand the functionality of the system for monitoring other types of personal protective equipment and integration with construction management information systems.