Evaluation of the Versatility of a Traffic Anomaly Detection Method during Heavy Rainfall

Abstract

AbstractDuring heavy rains, traffic monitoring is limited, as the affected areas are monitored by reporting and patrolling. In this study, a method for detecting traffic anomalies during heavy rainfall events was established, and a model that uses probe vehicle data to detect traffic anomalies during a disaster (an event in which vehicles make U-turns in front of a damaged area) was proposed. In addition, a parameter calibration method was developed for the model using past disaster-related data. The generalizability of the calibrated model was evaluated by applying it to other disasters. According to the results, the proposed model exhibited good generalizability.