The simulation models on pedestrian moving are of great importance to the study of pedestrian dynamics in subway stations. Due to the lack of high-precision pedestrian motion data, these improved models are rarely verified and not always reliable. The trajectories obtained by manually annotating the individual locations frame by frame are high-precision, but this method is difficult to perform when the population density is high. Based on the deep learning algorithm, this paper innovatively proposes a more efficient method on trajectory collection for pedestrians in subway stations. Firstly, by analyzing the characteristics of video on pedestrian flows in subway stations, the YOLOv4 is selected as the object detector and the Deep SORT algorithm is selected as the tracker for tracking pedestrians. Then the conversion process between pixel coordinates and world coordinates are introduced to obtain the trajectories in reality. To verify the proposed method, the observed trajectories are compared with the ground-truth trajectories. The result show that they were basically consistent and there has the MPE (Mean Position Error) ~ N (0.37, 0.0982). Finally, from the perspective of trajectory similarity and moving parameters, the SFM (Social Force Model) considering overtaking behavior is verified and proved to be more effective than the original and can reproduce the more realistic pedestrian moving process.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Method on Trajectory Collection for Pedestrians in Subway Station Based on Deep Learning Algorithm


    Additional title:

    Lect. Notes Electrical Eng.


    Contributors:
    Liang, Jianying (editor) / Jia, Limin (editor) / Qin, Yong (editor) / Liu, Zhigang (editor) / Diao, Lijun (editor) / An, Min (editor) / Shi, Yihan (author) / Xu, Jie (author) / Zhang, Hui (author)

    Conference:

    International Conference on Electrical and Information Technologies for Rail Transportation ; 2021 October 21, 2021 - October 23, 2021



    Publication date :

    2022-02-19


    Size :

    8 pages





    Type of media :

    Article/Chapter (Book)


    Type of material :

    Electronic Resource


    Language :

    English




    Method on Trajectory Collection for Pedestrians in Subway Station Based on Deep Learning Algorithm

    Shi, Yihan / Xu, Jie / Zhang, Hui | British Library Conference Proceedings | 2022


    Subway station

    YIN ZHAOFU / ZHANG BO / ZHOU TING et al. | European Patent Office | 2022

    Free access

    SYSTEM AND METHOD FOR TRAJECTORY PLANNING FOR UNEXPECTED PEDESTRIANS

    GUPTA RAKESH / COFIELD ROBERT | European Patent Office | 2017

    Free access

    Automated Pedestrians Data Collection Using Computer Vision

    Sayed, Tarek / Zaki, Mohamed / Tageldin, Ahmed | Springer Verlag | 2016


    Subway Station Induction System

    KIM EUI JIN / KIM YOUNG TAEK | European Patent Office | 2020

    Free access