The invention discloses an intelligent traffic multi-model detection method and system, and the method comprises the steps: replacing an original loss function with an SIOU, replacing a C3 backbone network with GhostNet, and employing 9-Mosaic to improve the performance during weight file training. During detection, a yov5 weight file is read firstly, parameters are set, and road signs such as zebra crossings, lane lines, traffic lights and the like are calibrated. When the vehicle is detected, the vehicle area is read and stored, the slices are input into the MTCNN network to recognize the license plate, if the license plate is recognized, the slices are input into the LPRNet character recognition network, and the license plate content is returned. When pedestrians are detected, the pedestrian area is read and stored, and the calibrated traffic light area slices are identified by the ResNet residual network. And comprehensively judging whether the vehicle or the pedestrian relates to an illegal behavior according to the data, and storing the important part as a subsequent judgment basis. A primary and secondary multi-model is adopted, the primary model undertakes primary detection work, and the secondary model undertakes scattered recognition.

    本发明公开一种智慧交通多模型检测方法及系统,在权重文件训练时,采用SIOU替换原损失函数、更换C3主干网络为GhostNet,使用9‑Mosaic以提升性能。检测时先读取yolov5权重文件,设定参数,标定斑马线,车道线,红绿灯等道路标志。检测车辆时读取并保存车辆区域,将切片输入MTCNN网络对车牌识别,若识别出车牌将切片输入LPRNet文字识别网络,返回车牌内容。检测行人时,读取保存行人区域,由ResNet残差网络对标定红绿灯区域切片做识别。将数据综合判断车辆或行人是否涉及违规行为并将重要部分保存作为后续判断依据。本发明采用主次多模型,主模型承担主要检测工作,次级模型承接零散识别。


    Access

    Download


    Export, share and cite



    Title :

    Multi-model detection method and system for smart traffic


    Additional title:

    一种面向智慧交通的多模型检测方法及系统


    Contributors:
    SUN JUNXIN (author) / YU JIAJIA (author) / SONG YIRUN (author) / WU ZEHUA (author) / ZHOU SIYU (author)

    Publication date :

    2023-06-23


    Type of media :

    Patent


    Type of material :

    Electronic Resource


    Language :

    Chinese


    Classification :

    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



    SMART TRAFFIC SYSTEM

    CHENG QI | European Patent Office | 2022

    Free access

    Smart traffic system and method

    ZOU LIJIAN / YU CHONGJUN / CONG LIN | European Patent Office | 2022

    Free access

    Smart Traffic System

    CHOI SUNG WOOK / KIM DONG HYUN | European Patent Office | 2021

    Free access

    Smart Traffic System: Detection of the Emergency Clearance

    Ariffin, Mohamad Safwan Mohamad / Khairam, Haziq / Seman, Mohamad Tarmizi Abu | Springer Verlag | 2022