The invention relates to a traffic anomaly detection method based on a graph convolutional neural network auto-encoder, and designs a one-dimensional convolution and context coding network combining traffic anomaly and deep learning, the network mainly comprises a mirror image time domain convolution module and two graph convolution gating circulation modules cascaded in sequence, traffic conditions and possibly sent anomalies are predicted by extracting traffic feature images, namely speed and flow, a self-adaption method is used before a mirror image time domain convolution module to adapt to different road sections, more features are transmitted to the time convolution module through mirror images, more information is obtained through the time convolution module, and the time convolution module is used for processing the traffic features. The network continuously learns the traffic network, a Gaussian kernel function module is used in a graph convolution gating circulation module, distribution is more concentrated in a high-dimensional space, hidden spatial correlation is captured by using the characteristics of a graph convolution network architecture, possible abnormal points are captured in combination with a graph convolution neural network, the method is more accurate, and the algorithm is more efficient. And the reliability of abnormal prediction is greatly improved.

    本发明涉及一种基于图卷积神经网络自编码器的交通异常检测方法,设计一种结合交通异常和深度学习的一维卷积及上下文编码网络,所述网络主要包括镜像时域卷积模块和依次级联的两个图卷积门控循环模块,通过提取交通特征像是速度和流量来预测交通状况和可能发送的异常,在镜像时域卷积模块之前使用了自适应方法来适应不同的路段,通过镜像传入更多的特征给时间卷积模块,通过时间卷积模块获取更多的信息,让网络不断地学习这种交通网络,图卷积门控循环模块使用了高斯核函数模块,让分布更加集中于高维空间,再利用图卷积网络架构的特点捕获了隐藏的空间相关性结合图卷积神经网络捕获可能的异常点发生,更加准确,大大提高了预测异常的可靠性。


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Traffic anomaly detection method based on graph convolutional neural network auto-encoder


    Weitere Titelangaben:

    基于图卷积神经网络自编码器的交通异常检测方法


    Beteiligte:
    LI XIAOJIE (Autor:in) / REN ZHIYU (Autor:in) / SHI CANGHONG (Autor:in) / WU XI (Autor:in) / CHEN KEN (Autor:in) / LYU JIANCHENG (Autor:in)

    Erscheinungsdatum :

    2023-09-19


    Medientyp :

    Patent


    Format :

    Elektronische Ressource


    Sprache :

    Chinesisch


    Klassifikation :

    IPC:    G06F ELECTRIC DIGITAL DATA PROCESSING , Elektrische digitale Datenverarbeitung / 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




    Aircraft brake system anomaly detection method based on generative adversarial network and auto-encoder

    ZHANG DONGPING / WANG MENGTING / XIA YUEJIAN et al. | Europäisches Patentamt | 2023

    Freier Zugriff

    Traffic jam prediction method based on tense graph convolutional neural network

    ZHANG HAO / ZHANG GE / HUA QIFAN et al. | Europäisches Patentamt | 2023

    Freier Zugriff

    An Auto-Encoder Based TinyML Approach for Real-Time Anomaly Detection

    Sai Charan, Kovuru | British Library Conference Proceedings | 2022


    Ship traffic flow prediction method based on improved graph convolutional neural network

    LIU WEN / ZHAN YANG / LIANG MAOHAN et al. | Europäisches Patentamt | 2022

    Freier Zugriff