AUTHOR=Yang Jun , Gong Mengjie , Dong Xueru , Liang Jiahua , Wang Yan 

TITLE=MPDNet: A Transformer-based real-time passenger detection network in metro stations

JOURNAL=Frontiers in Physics

VOLUME=Volume 10 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/physics/articles/10.3389/fphy.2022.1017951

DOI=10.3389/fphy.2022.1017951

ISSN=2296-424X

ABSTRACT=In order to achieve subway passenger flow detection in highly crowded scenes, this research proposed MPDNet, a passenger flow detection model for dense areas in metro stations. First, we implemented pyramid vision transformer to extract feature and global information, and then improved FPN with ASFF algorithm to enhance the local spatial information of features while feature fusion. In addition, to better evaluate the performance of models in metro, we collected subway surveillance video data and proposed the MetroStation dataset. The dataset contains the four main operating areas of the station, which better reflects the density of passenger flow and inter-passenger occlusion conditions. The experiments on the MetroStation dataset demonstrated that the MPDNet performed well and has satisfied inference speed in metro passenger flow detection.