Presentation: 2025 ND EPSCoR Annual conference 

October 21, 2025, NDSU Memorial Union, Fargo, North Dakota

Proactive Traffic Safety Assessment at High-Risk Intersections Using Machine Vision Applications

Traditional crash-based safety evaluations are reactive and limited by underreporting and delayed insights. Conflict analysis provides a proactive alternative by examining near-miss events, offering valuable insight into critical road user interactions and supporting proactive traffic safety assessments. This study introduces a conflict analysis framework that uses traffic CCTV footage and advanced video analytics to detect and classify intersection conflicts. Accurate trajectories were estimated with a custom object detection model, and Post-Encroachment Time (PET) and Time to Collision (TTC) were calculated based on those estimates for conflict identification. Peak over Threshold (POT) univariate Extreme Value Theory (EVT) and clustering methods, applied separately to PET and TTC, classify conflicts into low, medium, high, and very high severity categories. Manual conflict data was gathered by identifying events with PET ≤ 4s, which validated the accuracy of the automatic detection algorithm. The framework achieved a high accuracy of 99% in identifying different conflict types. Conflict severity analysis showed that PET and TTC values exceeding 1 second were consistently associated with low-severity outcomes, indicating that even a minimal buffer of 1 second usually provides drivers with enough time to perform evasive maneuvers and avoid critical interactions. Overall, this framework offers a scalable, plug-and-play solution for proactive safety monitoring using existing camera infrastructure. It addresses key gaps in prior studies by integrating detection, severity modeling, and behavior analysis, thereby supporting data-driven safety improvements aligned with Vision Zero goals.