Leveraging Distributed Acoustic Sensing for Vehicle Counting in High-Traffic-Density Environments

Urban traffic management poses a significant challenge for cities worldwide, often hampered by costly conventional infrastructures and devices such as cameras and magnetic loops. In this paper, we present an original approach to vehicle counting using Distributed Acoustic Sensing (DAS) technology. Leveraging existing telecommunication optical fibre cables along roadways, our method offers precise and high-resolution vehicle counting capabilities. Our approach employs a neural network model trained on a designated road segment, and we extend this model to other road segments through optimal transport-based feature mapping. The performance exhibits the adaptability of the model across diverse road sections and traffic scenarios, showcasing its robustness and potential for widespread deployment.