AUTHOR=Ye Zhipeng , Wang Weijun , Wang Xin , Yang Feng , Peng Fei , Yan Kun , Kou Huadong , Yuan Aijing 

TITLE=Traffic flow and vehicle speed monitoring with the object detection method from the roadside distributed acoustic sensing array

JOURNAL=Frontiers in Earth Science

VOLUME=Volume 10 - 2022

YEAR=2023

URL=https://www.frontiersin.org/journals/earth-science/articles/10.3389/feart.2022.992571

DOI=10.3389/feart.2022.992571

ISSN=2296-6463

ABSTRACT=Distributed Acoustic Sensing (DAS) is an emerging technique that converts a standard glass telecommunication fiber into an array of seismic sensors. Hence, DAS can simultaneously record the vibration of passing vehicles over tens of kilometers and thus has the potential to monitor traffic at low cost and low maintenance. However, automatically identifying and tracking vehicles on the road in real-time with big-data DAS recording is still full of challenges. Therefore, we propose a deep learning method based on the YOLOv5 object detection algorithm to estimate traffic flow and vehicle speed in DAS signals and evaluate them over a 500 m fiber segment in suburban Beijing. We created a dataset containing about 10,000 images with one-minute filtered DAS signals and labeled them manually. We randomly divided the dataset into training, validation, and testing sets in a ratio of 80%, 10%, and 10%. After about ten hours of training on a server with two Nvidia GeForce RTX 3090 GPUs, the precision reached 95.9%. We compared the performance with a beamforming technique on the same DAS data, and the results show that our method is much faster than the beamforming technique with equivalent performance. Further, we analyzed the temporal traffic trend of the road section and explored vehicle classification by weights.