Rapid growth in delivery and freight transportation is increasing in urban areas; as a result the use of delivery trucks and light commercial vehicles is evolving. Major cities can use traffic counting as a tool to monitor the presence of delivery vehicles in order to implement intelligent city planning measures. Classical methods for counting vehicles use mechanical, electromagnetic or pneumatic sensors, but these devices are costly, difficult to implement and only detect the presence of vehicles without giving information about their category, model or trajectory. This paper proposes a Deep Learning tool for classifying vehicles in a given image while considering different categories of logistic vehicles, namely: light-duty, medium-duty and heavy-duty vehicles. The proposed approach yields two main contributions: first we developed an architecture to create an annotated and balanced database of logistic vehicles, reducing manual annotation efforts. Second, we built a classifier that accurately classifies the logistic vehicles passing through a given road. The results of this work are: first, a database of 72 000 images for 4 vehicles classes; and second two retrained convolutional neural networks (InceptionV3 and MobileNetV2) capable of classifying vehicles with accuracies over 90%.


    Access

    Download


    Export, share and cite



    Title :

    Classifying logistic vehicles in cities using Deep learning


    Contributors:

    Publication date :

    2019


    Remarks:

    World Conference on Transport Research, May 2019, Mumbai, India


    Type of media :

    Preprint


    Type of material :

    Electronic Resource


    Language :

    English




    MONITORING OF LOGISTIC VEHICLES

    KUMAR AVISHEK / MOLNAR CHRISTIAN / SCHLOEMER JAMES F et al. | European Patent Office | 2021

    Free access

    Method for classifying moving vehicles

    LEHNING MICHAEL / PROEFROCK DIMA | European Patent Office | 2016

    Free access

    Method for classifying moving vehicles

    LEHNING MICHAEL / PROEFROCK DIMA | European Patent Office | 2017

    Free access

    Division 86 - Armored Combat Logistic Vehicles

    Russ,M.D. / US Army,US | Automotive engineering | 1982


    Classifying Road Intersections using Transfer-Learning on a Deep Neural Network

    Baumann, Ulrich / Huang, Yuan-Yao / Glaser, Claudius et al. | IEEE | 2018