The invention relates to a traffic accident identification method based on a time-space preorder coding and decoding predictor; the method comprises the following specific steps: carrying out the feature extraction through employing a parallel convolutional neural network, and extracting the time periodicity of traffic speed and flow, and the spatial correlation between upstream and downstream; and then ensuring synchronous capture of the spatial-temporal characteristics of the traffic flow through spatial-temporal position embedding, and capturing static and dynamic spatial-temporal dependencies of the traffic flow by combining the spatial-temporal characteristics of the traffic flow with a spatial-temporal Transform; and then, acquiring the accident occurrence probability by combining the traffic flow empty dependency during multiple extraction with the convolutional neural network, and finally, accurately identifying the occurrence of the traffic accident in real time. According to the method, urban traffic accidents can be effectively identified by utilizing urban traffic data such as fixed station average speed and fixed station average flow, effective information is provided for urban traffic safety, and the method has the characteristics of high identification precision, strong real-time performance, short model training time and the like.
本发明涉及一种基于时空前序编解码预测器的交通事故识别方法,具体步骤如下:使用并列的卷积神经网络进行特征提取,提取交通速度和流量的时间周期性和上下游的空间关联性,随后通过时空位置嵌入保证交通流时空特征的同步捕捉,并将交通流的时空特征结合时空Transformer捕获交通流静态和动态的时空依赖性。之后通过多次提取时交通流空依赖性结合卷积神经网络得到事故发生概率,最终实时地精准地识别交通事故的发生。本发明能够利用城市交通数据:固定站点平均车速,固定站点平均流量,对城市交通事故进行有效识别,为城市交通安全提供有效信息,具有识别精度高、实时性强、模型训练时间短等特点。
Traffic accident identification method based on time-space preorder coding and decoding predictor
一种基于时空前序编解码预测器的交通事故识别方法
2021-10-22
Patent
Electronic Resource
Chinese
Speed choice versus celeration behavior as traffic accident predictor
Online Contents | 2006
|Method for Receiving Preorder Goods by using a Cigar Jack Device
European Patent Office | 2017
|Complete machine pre-operation preorder safety monitoring method and system
European Patent Office | 2024
|