Mixed event detection with YOLOv9 based on distributed optical fiber acoustic sensing

Distributed Optical fiber sensing technology plays an important role in perimeter security due to its advantages such as long detection distance, high sensitivity, and flexible deployment. However, the problem of accurate detection is still faced in the detection of multiple vibration mixed events where multiple events occur at the same time. In this study, a method for detecting multiple vibration mixed events is proposed. This method was developed by integrating distributed acoustic sensing (DAS) technology with the YOLOv9 network. Vibration data from three events (walking, shaking, knocking) and their random combinations were acquired using DAS, followed by the conversion of the collected data into spatio-temporal maps, which were then processed by the YOLOv9 network for highly efficient and accurate image detection and event identification. In the results, the average precision, recall, and mAP50 can reach 0.95, 0.95, 0.98 in event detection. It was confirmed that the integrated DAS and YOLOv9 method provided reliable technical support for efficient and precise detection of multi-vibration mixed events and demonstrated significant potential for application in perimeter security.