Lane detection is considered one of the most vital autonomous driving systems, which can guide vehicles in a completely safe and efficient manner. Painted lane markings based conventional lane detection systems malfunction more often in roads which do not have markings, commonly found in rural or less-developed areas. This deficiency impedes the general adaptation of autonomous vehicle technologies. We present an advanced system, ROADCEPTION, that combines lane detection with deep learning architectures for object recognition. The system is achieved by leveraging Convolutional Neural Networks (CNN) for road boundary identification, and YOLO that allows for real-time objects such as vehicles, pedestrians, and traffic signs, enabling it to overcome the problem of unstructured road environments. ROADCEPTION combines advanced edge detection and region-of-interest analysis techniques to maximize performance in real-time. Its capabilities are robust enough for safe and reliable operation under road conditions with faded lane markings or dynamic obstacles. In extensive evaluations on various datasets, the system has shown accurate performance in lane detection as well as object recognition with very low latency, making it fit for real-world applications. It empowers autonomous vehicles with the ability to navigate environments that have minimal road infrastructure with safety and efficiency. By adapting to complex driving scenarios, such as rural terrains and poorly maintained infrastructure, ROADCEPTION addresses critical limitations of traditional lane detection methods. Its scalable and versatile approach significantly enhances the safety and reliability of autonomous navigation across various geographic and environmental contexts with an accuracy of $84.86 \%$. As a pivotal advancement in autonomous vehicle technology, ROADCEPTION bridges the gap between conventional algorithms and the real-world demands of modern transportation systems. Furthermore, the system’s modular design ensures ease of integration with existing autonomous driving platforms.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    ROADCEPTION: Intelligent Lane Identification and Object Detection for Enhanced Road Safety in Autonomous Vehicles


    Beteiligte:


    Erscheinungsdatum :

    28.04.2025


    Format / Umfang :

    708066 byte




    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch



    Object identification for autonomous road vehicles

    NICKEL JANICE H / NICKEL BRYAN H | Europäisches Patentamt | 2021

    Freier Zugriff

    Object identification for autonomous road vehicles

    NICKEL JANICE H / NICKEL BRYAN H | Europäisches Patentamt | 2025

    Freier Zugriff

    Object identification for autonomous road vehicles

    NICKEL JANICE H / NICKEL BRYAN H | Europäisches Patentamt | 2021

    Freier Zugriff

    Implementation of Road Safety Perception in Autonomous Vehicles in a Lane Change Scenario

    Del Re, Enrico / Olaverri-Monreal, Cristina | ArXiv | 2022

    Freier Zugriff