Incident Detection Method using Longitudinal Occupancy Time-Series Data

抄録

Incidents frequently occur in the expressway. A fast and precise detection of incidents is required to mitigate negative impacts caused by delay of traffic managements. This study proposes an incident detection method using a non-parametric model. In the proposed method, traffic incidents are detected by developing a conditional probability function of traffic state using the long term data which is observed by traffic detectors (longitudinal occupancy time-series data). The proposed method was verified empirically using actual field data, then compared with existing incident detection methods. Analysis results show that the proposed method has high applicability due to no need of complex parameter calibrations.