Abstract
This study focuses on the development of a multisensory unmanned aerial vehicle (UAV) system for automatic detection of foreign objects on transport infrastructure. The research addresses the limitations of traditional monitoring methods, such as low detection accuracy and limited coverage, by integrating optical, thermal, and radar sensors with advanced AI models (YOLOv8x and RPN+STN). The developed system demonstrated a detection accuracy of 98% and a detection range of up to 350 m during laboratory tests, while field trials on a 5 km highway and 2 km railway section achieved an overall accuracy of 92.0% and a false positive rate of 4.1%. These results are explained by the synergistic combination of multisensory data fusion and real-time AI processing, enabling robust detection in diverse environmental conditions. The distinctive feature of the proposed system is its ability to operate with high precision and low latency (15 FPS, 120-250 ms) in real-time scenarios. The system is suitable for practical deployment in road and railway monitoring, particularly for improving transport safety in regions with limited human resources and challenging terrain.