| Abstract |
Phase-sensitive Optical Time Domain Reflectometry (Phi-OTDR) systems are crucial for infrastructure monitoring and security, but face challenges in accurately and efficiently detecting real-time disturbance events. This study aims to enhance the real-time detection performance of four critical events (striking, excavating, traffic interference, and road breaking) in Phi-OTDR systems. To address this, we innovatively apply the YOLOv9 network to Phi-OTDR event real-time detection, incorporating Programmable Gradient Information (PGI) and Generalized Efficient Layer Aggregation Network (GELAN). Experimental results demonstrate that YOLOv9 outperforms previous methods across all evaluation metrics and event types, with notable improvements in precision (7.4% for road breaking and 2.1% for striking events) and recall rates (11.4% for traffic interference and 10.9% for road breaking events). YOLOv9 also exhibits significant advantages in the mAP@50 and mAP@50-95 metrics, particularly in detecting excavation and road breaking events. These improvements enhance the detection capabilities of Phi-OTDR systems, providing more reliable technical support for smart city development and infrastructure protection. © 2025 SPIE. |
