3D object detection from monocular camera videos constitutes a critical research domain. Achieving robust 3D object detection in databases lacking annotated information poses a highly challenging task. This paper proposes a simple yet effective transfer learning approach, integrating data alignment, 3D object detection, and dynamic result correction. Vanishing point detection is employed to infer camera angles in diverse scenes, and preprocessing of new data is conducted by considering the camera's pitch angle and vanishing point position. Subsequently, MonoDETR are applied for depth estimation and 3D object detection in monocular videos. Building upon the detection results, dynamic correction is achieved through inter-frame assistance, culminating in the final 3D object information. Validation on the TUAT Near-Miss Incident Database demonstrates the efficacy of the proposed approach. The results indicate a substantial reduction in the cost of annotating new domain data while simultaneously enhancing detection accuracy and robustness. Integration with onboard diagnostics (OBD) data allows the reconstruction of information about various traffic participants in hazardous scenarios, providing valuable insights for in-depth analysis of accident causes.


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Robust 3D On-Road Object Detection and Distance Estimation for Active Vehicle Control Systems Based on Monocular Camera Image Data


    Weitere Titelangaben:

    Lect.Notes Mechanical Engineering


    Beteiligte:
    Mastinu, Giampiero (Herausgeber:in) / Braghin, Francesco (Herausgeber:in) / Cheli, Federico (Herausgeber:in) / Corno, Matteo (Herausgeber:in) / Savaresi, Sergio M. (Herausgeber:in) / Zhang, Xingguo (Autor:in) / Ikami, Daiki (Autor:in) / Raksincharoensak, Pongsathorn (Autor:in)

    Kongress:

    Advanced Vehicle Control Symposium ; 2024 ; Milan, Italy September 01, 2024 - September 05, 2024



    Erscheinungsdatum :

    04.10.2024


    Format / Umfang :

    7 pages





    Medientyp :

    Aufsatz/Kapitel (Buch)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch





    Distance estimation and vehicle position detection based on monocular camera

    Ali, Abduladhem Abdulkareem / Hussein, Hussein Alaa | IEEE | 2016


    Active Learning based Robust Monocular Vehicle Detection for On-road Safety Systems

    Sivaraman, S. / Trivedi, M.M. | British Library Conference Proceedings | 2009


    Active learning based robust monocular vehicle detection for on-road safety systems

    Sivaraman, Sayanan / Trivedi, Mohan Manubhai | IEEE | 2009


    A Hybrid Framework for Object Distance Estimation using a Monocular Camera

    Patel, Vaibhav / Mehta, Varun / Bolic, Miodrag et al. | IEEE | 2023


    Robust Estimation of Vehicle Ego-motion for Monocular Camera System

    Koitabashi, H. / Iketani, T. / Murata, Y. et al. | British Library Conference Proceedings | 2008