A Real-Time Proactive Intersection Safety Monitoring System Based on Video Data

Status: Completed

Funding Agency: U.S. Department of Transportation University Transportation Center Program

Abstract:

Road user behavior and conflicts at intersections have become a working data source for traffic safety evaluation. This project built an AI video analytic tool that assesses intersection safety through surrogate safety measures and non-compliance behavior, and a method for ranking intersections by safety. Post-encroachment time and time-to-collision were used to identify rear-end and left-turning conflicts. Trajectory data came from a YOLO-v5 detection model integrated with a DeepSORT tracking framework, and extreme value theory was used to estimate crash counts at each intersection from the frequency of those measures.

Investigators: Dr. Mohammad Jalayer, PhD., Dr. Nidhal Carla Bouaynaya, PhD.

Links to Publications: 

A Real-Time Proactive Intersection Safety Monitoring System Based on Video Data, CAIT-UTC-REG53