Technology for Automated Monitoring of Construction Object Using Aerial Photography and Neural Networks
Keywords: Technical supervision, Aerial photography, Unmanned aerial vehicles (UAVs), Neural networks, Building inspections, Machine learning
Abstract. The quality of finished construction products is often compromised by various negative factors, including flawed engineering decisions, unqualified labor, and budget constraints. To mitigate these issues, technical supervision is essential for ensuring adherence to project and regulatory standards regarding timelines, costs, and quality. Traditional methods of technical supervision rely heavily on manual measurements and site inspections, necessitating the use of high-precision geodetic equipment operated by specialists. This research introduces an automated technology for monitoring the actual positions of capital construction objects through aerial photography captured by unmanned aerial vehicles (UAVs) and enhanced by neural networks. It aims to monitor structural geometry by comparing measured coordinates with design specifications. The proposed methodology is exemplified through the monitoring of pile positions in foundation construction, emphasizing the economic advantages of UAV-based monitoring over traditional geodetic methods, particularly in complex soil conditions. By utilizing calibrated UAV-mounted cameras and ensuring appropriate image overlap, this approach enhances the accuracy and reliability of coordinate measurements, ultimately contributing to improved construction quality and compliance with regulatory standards.