Most of the studies related to pedestrian behavior focus on signalized intersections while a few focus on overpasses. In this study pedestrians’ behaviors at overpass locations are investigated. For this purpose, two overpass locations were selected and pedestrians either using the overpass or crossing illegally were recorded by video cameras and 2713 accepted/rejected time gaps were measured. Raff’s method (deterministic approach) is used to estimate the critical gap. The collected data were then evaluated by a binary logit model (probabilistic approach) to estimate time gaps. The selected overpasses were observed on weekdays continuously for six hours. At the overpass locations 656 illegal crossings were observed on weekdays continuously for six hours. Safety margin of 377 pedestrians were evaluated by ANOVA analysis to identify the significant factors. ANOVA results revealed that interactions of gender, age, and vehicle position, items carrying, group size had significant effects on safety margin.