Vision based traffic conflict analysis

In this paper a vision based system for traffic conflict analysis is presented. The system employs geo-referenced stereo sequences and tracking procedure to provide much greater information regarding pre-crash and crash events than what is currently available. The exploitation of the traffic conflict technique potentiality as a surrogate safety measure could constitute an effective tool in understanding how the driver interacts and adapts its behavior with respect to the vehicle, the road characteristics, the traffic control devices and the environment. Experiments performed on real data acquired in urban context confirm the effectiveness of the system for the traffic conflicts measurement and the driver behavior analysis.