Light detection and ranging sensors (LIDAR) play a fundamental role in the perception systems of Unmanned Ground Vehicles. However, particulate matters existing in smog scenarios such as fire disasters seriously degrade the robustness of LIDAR-based algorithms. Therefore, it’s essential for UGV to recognize and locate smog. Based on the Class Activation Map (CAM), we propose a lightweight smog detection model, where an active annotation method combining with activation loss is designed for enhancing the quality of CAM. Experiments demonstrate the effectiveness of the proposed method with just 4ms time costs on GTX-1050ti.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Lightweight Smog Detection Model for Unmanned Ground Vehicle Based on Interpretability of Neural Networks


    Weitere Titelangaben:

    Lect. Notes Electrical Eng.


    Beteiligte:
    Wu, Meiping (Herausgeber:in) / Niu, Yifeng (Herausgeber:in) / Gu, Mancang (Herausgeber:in) / Cheng, Jin (Herausgeber:in) / Sun, Yi (Autor:in) / Hu, Xiaochang (Autor:in) / Xu, Xin (Autor:in) / Li, Jian (Autor:in)

    Kongress:

    International Conference on Autonomous Unmanned Systems ; 2021 ; Changsha, China September 24, 2021 - September 26, 2021



    Erscheinungsdatum :

    18.03.2022


    Format / Umfang :

    10 pages





    Medientyp :

    Aufsatz/Kapitel (Buch)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch





    Lightweight Smog Detection Model for Unmanned Ground Vehicle Based on Interpretability of Neural Networks

    Sun, Yi / Hu, Xiaochang / Xu, Xin et al. | British Library Conference Proceedings | 2022


    Lightweight and Efficient Air-to-Air Unmanned Aerial Vehicle Detection Neural Networks

    Wang, Chuanyun / Li, Zhenfei / Gao, Qian et al. | IEEE | 2023



    Lightweight Trustworthy Message Exchange in Unmanned Aerial Vehicle Networks

    Liu, Zhiquan / Guo, Jingjing / Huang, Feiran et al. | IEEE | 2023