The Japanese government plans to replace the manual survey method of counting vehicles with CCTV images in order to obtain traffic volume by vehicle types automatically. The current method of manually checking vehicles and counting them as small or large vehicles is time-consuming, and there is a shortage of surveyors. Automating this process will allow for more accurate and constant monitoring of traffic volume. Existing research can be found in developing vehicle detection and tracking models using machine learning. However, these models have not yet achieved the accuracy required for practical use. This paper aims to develop a vehicle classification and tracking model with high accuracy for 7 classes of vehicle types. YOLOv5 is used for vehicle detection and classification, along with DeepSORT for vehicle tracking. The results showed most of the classes reached 95% accuracy both for classification and tracking.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Development of Robust Vehicle Classification and Tracking for Traffic Analysis


    Contributors:


    Publication date :

    2022-11-11


    Size :

    681583 byte




    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English



    Robust Vehicle Detection and Tracking for Traffic Surveillance

    Jung, Y.-K. / Ho, Y.-S. | British Library Conference Proceedings | 1999



    Vehicle detection, tracking and classification in urban traffic

    Chen, Zezhi / Ellis, Tim / Velastin, Sergio A | IEEE | 2012


    Automatic traffic surveillance system for vehicle tracking and classification

    Jun-Wei Hsieh, / Shih-Hao Yu, / Yung-Sheng Chen, et al. | IEEE | 2006