Development of Vision-based Driving Strategies for Traffic Junction Scenarios

Ee Heng Chen

Abstract: Crossing a traffic junction is a challenging task for autonomous vehicle. It must first decide if the traffic junction is safe to be crossed, which depends on factors such as the presence of stop signs, the state of traffic lights, and the behavior of traffic participants approaching the traffic junction. In this thesis, a vision-based system is introduced to carry out the decision-making process. Experiments performed using synthetic and real-world data show the feasibility of the system in inferring the right decision based solely on information extracted from camera images.