Measuring the Severity of Violations Committed by Pedestrians at Signalized Intersections

Abstract

Pedestrian safety remains a global concern, with pedestrians accounting for a substantial proportion of road fatalities and injuries. Previous research has examined many factors that influence pedestrian safety however, pedestrian violations at signalized intersections are a major contributor to pedestrian-vehicle conflicts, yet existing traffic safety research primarily treats these behaviours as binary outcomes, either as compliant or non-compliant. This representation fails to capture the varying degrees of severity in different crossing paths, limiting a nuanced understanding of pedestrian risk and road safety. To address this gap, this thesis proposes, applies, and evaluates continuous metrics to characterize the severity of both spatial and temporal pedestrian violations at signalized intersections. Three spatial violations (SV) metrics and four temporal violation (TV) metrics were developed. To evaluate these measures, observational pilot studies were conducted at a highly active signalized intersection in Waterloo, Ontario, utilizing both drone and street-level video data. The spatial violation analysis revealed that 48% of pedestrians committed a spatial violation, with the vast majority (86.5%) engaging in corner cutting. Among the spatial metrics, SV Metric 2 (total violation area) and SV Metric 3 (time-step metric) provided the most robust and consistent severity evaluations, whereas SV Metric 1 (maximum lateral distance) tended to overestimate spatial violation severities. The temporal violation analysis found that 7.6% of pedestrians committed a temporal violation, predominantly initiating crossings during the Flashing Don’t Walk interval (86.7%). Unlike the spatial metrics, the temporal metrics yielded statistically distinct severity distributions, demonstrating that the measurement of temporal violation severities is highly sensitive to whether the selected metric prioritizes strict legal compliance or practical safety risk. Ultimately, this research recommends the use of SV Metrics 2 and 3 for spatial analysis and provides a framework for selecting appropriate TV metrics based on specific study parameters. The proposed metrics can be integrated into existing automated traffic data collection systems, supporting future research relating violation severity to conflict severity.