In object perception for automated driving, Oriented Bounding Box(OBB) detection is important for extracting the position and direction of objects. However, previous studies on OBB detection dealt only with general situation such as L-shaped objects, and have lower performance for other different shaped objects. Thus, we focused on various shape such as U-shaped, Complex-shaped, and other non L-shaped objects that appear on road. In this paper, we propose a new OBB detection method robust to several challenging shapes and efficient to implement under realtime constraints. The proposed method encodes 3D points of each object into 2D grids, leading to the discretization effect that helps the rectangle fit the object. The oriented angle is extracted through our angle optimization approach using minimum loss solution based on the distance between the grid and the rectangle. Experiments were conducted with real-world data obtained with LiDAR and Differential-GPS(DGPS), and the results demonstrate the robustness and efficiency of the proposed method.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Oriented Bounding Box Detection Robust to Vehicle Shape on Road Under Real-Time Constraints


    Beteiligte:
    Baeg, Juheon (Autor:in) / Park, Jongwon (Autor:in)


    Erscheinungsdatum :

    24.09.2023


    Format / Umfang :

    2961177 byte





    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch




    BEYOND BOUNDING BOXES: USING BOUNDING SHAPES FOR REAL-TIME 3D VEHICLE DETECTION FROM MONOCULAR RGB IMAGES

    Gählert, Nils / Wan, Jun-Jun / Weber, Michael et al. | British Library Conference Proceedings | 2019


    Bounding box splitting for robust shape classification

    Bauckhage, C. / Tsotsos, J.K. | IEEE | 2005


    Bounding Box Splitting for Robust Shape Classification

    Bauckhage, C. / Tsotsos, J. K. | British Library Conference Proceedings | 2005