Roussou S., Ventura R., Ziakopoulos A., Yannis G., “Predicting Pedestrian Violations in Urban Intersections Using ANOVA and Poisson Regression”, Proceedings of the 12th International Congress on Transportation Research (Thessaloniki, Greece, 16-18 October 2025).

This study investigates the factors influencing pedestrian violations at urban intersections such as Panepistimiou Street & Omonoia Square and Panepistimiou Street & Leoforos Vasilissis Sofias Street in Athens, Greece. Using data collected through computer vision techniques (YOLOv8 for object detection, ResNet-50 for feature extraction, and Kalman filtering for trajectory tracking), the study applies statistical modeling techniques to analyze illegal pedestrian crossings. To investigate the statistical significance of key influencing factors, a two-way ANOVA was employed to find differences in pedestrian violation rates between the two different locations. In addition, Poisson regression was employed to model the likelihood of pedestrian violations. Results indicate that Time-To-Collision and pedestrian volume significantly affect crossing behavior as they influence non-compliance. The study provides a more complete explanation of the trends in pedestrian compliance, which can help city planners and policymakers create more effective interventions to strengthen road safety.