Collision Risk Assessment between U-turn and On-coming Vehicles by Trajectory Extraction from UAV

Abstract

Collisions between U-turn vehicles and oncoming vehicles have become a problem on intercity national highways in Thailand. The objective of this study is to develop an image processing technique to extract trajectories from videos taken by UAVs, and to quantitatively express and evaluate collision risks. In Thailand, many behaviors were observed in which the oncoming vehicle decelerates after recognizing a U-turn vehicle. Since the collision risk changes significantly depending on whether the oncoming vehicle decelerates or not, this study proposed an indicator to express these behaviors. The indicators were proposed to express the collision risk caused by a U-turn vehicle not slowing down and the collision risk caused by the oncoming vehicle misjudging the start of the U-turn. The combination of the indices showed the frequency of collision risk by vehicle type and by lane.