基于尖点突变理论模型的交通事故检测

Traffic incident detection based on the cusp catastrophe theory model

为了准确、及时地对交通事故进行自动检测,提出一种基于尖点突变理论的交通事故离群挖掘算法。首先,考虑到基于尖点突变理论描述交通流变化规律的可行性和有效性,从系统论的观点出发,基于尖点突变理论建立城市道路交通流表征模型,能够客观全面地反映交通拥挤条件下符合交通实际工况的速度、流量之间的数值关系,为交通事故离群挖掘奠定模型基础。然后,根据聚类原理,设计基于城市道路交通流表征模型的离群事件挖掘算法,从而检测获得表征交通事故的离群点,完成交通事故检测。最后,仿真算例验证所提出的交通事故检测算法具有较高的准确性,能有效区分常规拥堵与交通事故。

In order to perform automatic detection for the traffic incident accurately and timely, an outlier mining algorithm of traffic incident is proposed based on the cusp catastrophe theory. First of all, since describing traffic flow variation using the cusp catastrophe theory is feasible and effective, from the perspective of system theory, a representation model for urban road traffic flow established based on cusp catastrophe theory could objectively and comprehensively depict the numerical relation between vehicle speed and volume in congested traffic conditions, which creates the theoretical basis for the outlier mining of traffic incidents. Then, based on the clustering principle, an outlier mining algorithm was designed based on the established representation model of urban road traffic flow. The detection of accidents can be accomplished by identifying the outliers which characterize the traffic accidents. Finally, through simulations, the proposed algorithm was proved to be highly accurate, and it is capable of efficiently distinguishing the regular congestion from the accidents.