Processing video from urban video surveillance cameras requires the use of algorithms for multi-object detection and tracking on video in real-time. However, existing computer vision algorithms require the use of powerful equipment and are not sufficiently optimized to process multiple video streams simultaneously. This article proposes an approach to using the tracker in conjunction with the YoloV4 object detector for real-time video processing on medium-power equipment. Paper also presents the solution for difficulties that arise during work with optical flow. The results of the comparison of the accuracy and speed of image processing of the applied approach with such trackers as IOU17, SORT, KCF, and MOSSEE are also presented.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Real Time Multi Object Detection & Tracking on Urban Cameras


    Additional title:

    Sustain. Civil Infrastruct.



    Conference:

    International Road Federation World Meeting & Exhibition ; 2021 ; Dubai, United Arab Emirates November 07, 2021 - November 10, 2021



    Publication date :

    2022-04-21


    Size :

    12 pages





    Type of media :

    Article/Chapter (Book)


    Type of material :

    Electronic Resource


    Language :

    English




    High Precision Real-time 3D Tracking Using Cameras

    Mannberg, Mikael / Silson, Peter / Tsourdos, Antonios et al. | AIAA | 2011


    High Precision Real-time 3D Tracking Using Cameras

    Mannberg, M. / Silson, P. / Savvaris, A. et al. | British Library Conference Proceedings | 2011


    Situation analysis and atypical event detection with multiple cameras and multi-object tracking

    Reulke, Ralf / Meysel, Frederik / Bauer, Sascha | Tema Archive | 2008