Abstract
The safety assessment of signalized intersections is usually carried out by developing statistical and data-driven methods based on data aggregated at some temporal and spatial level (e.g., yearly or hourly; interval or approach segment). However, such aggregations suffer from significant simplifications that obscure patterns of potential spatiotemporal security risks within the level of data aggregation.The existing alternative indicators, Surrogate Safety Measures (SSMs), cannot effectively describe the real-time collision risk process of cross conflicts. This research presents a vehicle collision risk model based on fuzzy logic to express the real-time collision risk process of intersection conflicts in signalized intersections and proposes a function analysis method to perform a new characterization of the safety risk mode of signalized intersections.A signalized intersection with double right-turn lanes is selected as the research site, and about 1 hour of traffic video data recorded by a UAV is used to extract vehicle trajectory data. The proposed risk model is based on the characterization of the collision risk function, which can reveal the time distribution cycle of the risk of each lane within the signal. There is a difference between the same turning traffic movements of the opposite direction in the same phase. It is necessary to distinguish between two types of traffic movement in the opposite direction in the same phase.The proposed model has the potential to be used to facilitate active safety management and optimize signal timing while considering traffic safety on a more granular level.