AI and CCTV Analyze Soft Traffic Interventions in Minneapolis

A new study from arXiv introduces an AI-enabled analytics framework that uses existing CCTV infrastructure to evaluate the impact of soft infrastructure interventions on vehicle speed and safety. The research, conducted in Minneapolis, employed deep learning and perspective-based speed estimation to analyze driver behavior before and after installing temporary pedestrian refuges and curb extensions. At unsignalized intersections, mean speeds dropped by up to 18.75% and 85th-percentile speeds by 16.56%, with pass-through traffic decreasing by 12.2%. Signalized intersections saw reductions of up to 20.0% in mean speed and 17.19% in 85th-percentile speed, except at one location. The findings demonstrate the traffic-calming effectiveness of soft infrastructure, offering a scalable, low-cost method for urban design evaluation.