The invention discloses a multi-scale vehicle target detection method suitable for a highway monitoring scene, and belongs to the field of intelligent traffic. According to the vehicle target detection method, a YOLOv8 backbone network is improved on the basis of a ConvNeXt V2 network, and an SPD-Conv module and an SA-Conv module are introduced into the ConvNeXt V2 network, so that the multi-scale feature expression capability of a model is enhanced; secondly, a feature pyramid network HS-FPN is adopted to carry out fusion enhancement on a feature map extracted by the backbone network so as to cope with multi-scale detection of a vehicle target; and finally, carrying out classification and regression on the enhanced multi-scale feature map by adopting a detection head with a classification task and a regression task being mutually decoupled, and finally obtaining a detection result of the vehicle target. According to the method, the accuracy of multi-scale vehicle target feature extraction of the highway monitoring scene can be effectively improved, the conditions of large target scale change, many interference factors and the like in the scene can be better handled, and the detection accuracy can be improved while the detection efficiency is ensured.

    本发明公开了一种适于高速公路监控场景的多尺度车辆目标检测方法,属于智能交通领域。本发明车辆目标检测方法,以ConvNeXt V2网络为基础对YOLOv8骨干网络进行改进,并向ConvNeXt V2网络中引入SPD‑Conv模块和SA‑Conv模块,用于强化模型的多尺度特征表达能力;其次,采用特征金字塔网络HS‑FPN对骨干网络提取到的特征图进行融合增强,以应对车辆目标的多尺度检测;最后,采用分类任务和回归任务相互解耦的检测头对增强的多尺度特征图进行分类与回归,最终得到车辆目标的检测结果。本发明能够有效提高高速公路监控场景的多尺度车辆目标特征提取的精确度,更好地应对场景中目标尺度变化大、干扰因素多等状况,可在保证检测效率的同时提升检测精度。


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Multi-scale vehicle target detection method suitable for highway monitoring scene


    Weitere Titelangaben:

    适于高速公路监控场景的多尺度车辆目标检测方法


    Beteiligte:
    ZHAO MIN (Autor:in) / SUN DIHUA (Autor:in) / QIU ZUOWEI (Autor:in) / WANG SHISEN (Autor:in) / CHEN XINGZHOU (Autor:in) / WANG JINGTING (Autor:in) / ZHANG HAIBO (Autor:in)

    Erscheinungsdatum :

    16.07.2024


    Medientyp :

    Patent


    Format :

    Elektronische Ressource


    Sprache :

    Chinesisch


    Klassifikation :

    IPC:    G06V / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen / G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS



    Radar and camera detection target matching method and system in highway scene

    YANG YANG / CHEN XINKAI / CHEN KEN et al. | Europäisches Patentamt | 2025

    Freier Zugriff

    Vehicle control method and system used in highway scene

    ZUO GUOMIAO / LEI YONGFU / CHUNG WON-JIN et al. | Europäisches Patentamt | 2023

    Freier Zugriff

    Target vehicle track prediction method for multi-lane scene

    ZUO ZHIQIANG / GUO SONGLIN / WANG YIJING et al. | Europäisches Patentamt | 2023

    Freier Zugriff

    Vehicle multi-scene confluence control system and method in highway ramp confluence area

    YAN YING / ZHANG YU / ZHANG ZEYI et al. | Europäisches Patentamt | 2024

    Freier Zugriff

    Multi-target detection method for complex traffic scene

    LIN JUNJIE / ZHANG ZHOU / DENG WEI et al. | Europäisches Patentamt | 2025

    Freier Zugriff