The road network is essential for autonomous driving vehicles to navigate in intelligent transportation systems. With the success of deep learning, many methods regard road extraction as dense pixel-wise segmentation based on neural networks. However, the predicted fragmented road regions, resulting from frequently occurring occlusions, are challenging to utilize in topology-required autonomous driving applications. In this paper, we formulate road extraction as sparse anchor-points estimation and their connectivity construction problems to enhance the global topology of roads. We design a multi-task architecture within an encoder-decoder framework to simultaneously predict the segmentation, anchor-points and connectivity maps. Besides, we propose a topo-swin block and incorporate multi-level feature maps in the decoder to combine local information with long-distance topological relationships. Furthermore, a graph simplification criterion is designed to make it more efficient for autonomous driving applications. Experiments on the challenging Deepglobe dataset demonstrate that our proposed model outperforms existing road extraction methods by achieving 77.24% in topology-aware metrics.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Topology-aware Road Extraction via Multi-task Learning for Autonomous Driving


    Beteiligte:
    Li, Tao (Autor:in) / Ye, Shanding (Autor:in) / Li, Ruihang (Autor:in) / Fu, Yongjian (Autor:in) / Yang, Guoqing (Autor:in) / Pan, Zhijie (Autor:in)


    Erscheinungsdatum :

    24.09.2023


    Format / Umfang :

    1849201 byte





    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch




    Context-Aware Multi-Task Learning for Traffic Scene Recognition in Autonomous Vehicles

    Lee, Younkwan / Jeon, Jihyo / Yu, Jongmin et al. | IEEE | 2020



    Off-road autonomous driving

    LU JIANBO / HROVAT DAVOR DAVID / TSENG HONGTEI ERIC | Europäisches Patentamt | 2019

    Freier Zugriff

    Off-road autonomous driving

    LU JIANBO / HROVAT DAVOR DAVID / TSENG HONGTEI ERIC | Europäisches Patentamt | 2017

    Freier Zugriff