Instantaneous traffic near-miss identification based on vehicle speeds and proximity at signalized intersections

Traffic near misses refer to situations in which conflicting vehicles or vulnerable road users (VRUs) are about to crash but take evasive actions to avoid collision. Frequent near misses are strongly associated with crash potential, making them useful proactive safety indicators. Existing trajectory-based near-miss identification methods commonly require continuous tracking and post-processing of road-user trajectories, which can be computationally intensive for real-time applications. This paper presents a proximity-and-speed-based method for instantaneous near-miss identification at signalized intersections. The proposed framework scopes conflict zones by translating TTC/PET-based safety logic into spatial detection zones using vehicle speed, perception-reaction time, deceleration assumptions, and intersection geometry. A near miss is identified when conflicting fast vehicles occupy the scoped conflict zone during the relevant signal phase, allowing isolated event detection rather than continuous pairwise trajectory processing. The framework was tested at a LiDAR-equipped signalized intersection in Salt Lake City, Utah, USA. In a four-hour field evaluation, the algorithm identified 38 instantaneous near-miss events between permissive northbound left-turn vehicles and opposing southbound through vehicles using a 2.0-s post-encroachment time (PET) threshold. All LiDAR-reported events were validated through live monitoring and video review, and manually reviewed PET values were highly correlated with LiDAR-reported PET values. These findings demonstrate that the proposed conflict-zone-based framework can generate movement-specific and phase-specific near-miss events in real time while reducing the need for continuous trajectory processing of all vehicles.