An analysis system of pedestrian-vehicle interaction risk level using drone videos

Abstract

In Korea, out of the total number of car accidents between vehicles and pedestrians in 2015, the percentage of illegal crossing accidents is 40%. Unsignalized pedestrian crossing accidents is 38.8% of all accidents. Previously risk areas were determined by the occurrence and situation of accidents. This research proposes a collection of drone videos of general vehicle-pedestrian crossing behaviors to reflect on an analysis system to apply on assessing the risks of road safety. Firstly, acquiring drone video footage in selected research locations where vehicle and pedestrian risk level is sought. Secondly, with the footage collected going through a detecting and tracking process with each vehicle and pedestrian, within the footage collection period, labelling the vehicle and pedestrian positions, coordinates and speed. Thirdly, analyzing the vehicle and pedestrian interaction by use of Pedestrian Safety Margin (PSM) which is calculating the difference between the times a pedestrian crosses the conflict point at the time the next vehicle arrives at the same conflict point.