Innovative AI Video Analysis of Dilemma Zone Conflicts at Signal-controlled Intersections using Edge Computing and 5G

To address the critical needs of protecting road users from harmful traffic conflicts, a project is proposed for innovative AI video analysis of dilemma zone conflicts at signal-controlled intersections using edge computing and 5G. Specifically, the proposal applies innovative AI, advanced video analysis, and deep learning technologies, which show promising preliminary results, for automatic and robust analysis of dilemma zone conflicts and trajectory conflicts at intersections. For dilemma zone, videos covering 200-600 feet are analyzed to track vehicle behaviors for lane changing, speeds, types of vehicles, and number of vehicles when traffic lights change from green to yellow. For trajectory conflicts, videos covering the intersection area are analyzed to detect and log vehicles, pedestrians, cyclists crossing the intersection, and different types of trajectory conflicts, such as vehicle-to-vehicle (V2V), vehicle-to-pedestrian (V2P), vehicle-to-bicycle (V2B), and bicycle-to-pedestrian (B2P). The proof-of-concept of this phase I project will be expedited due to the success of the team's multiple related research on video analysis of traffic videos, as well as the promising preliminary results of the technical objectives, such as robust and accurate vehicle counting and traffic incidents detection.