Context-Enhanced Vehicle Tracking Method Under the Connected Environment (Poster)

Vehicle tracking is one of the key technologies for intelligent vehicle, which can prevent collision and save lives. The on-board sensors are commonly used for vehicle tracking while they always contain some noises and clutter measurements. The context-based information can bring several advantages for refining estimations, which thereby improves the performance of tracking. This paper proposes a novel context-enhanced vehicle tracking method with contextual information for modeling interaction behaviors between the vehicles and the environment, while the traditional algorithms assume that vehicles move independently. The approach combines interaction force with dynamic Bayesian networks by using context about the environment, object and traffic. In order to handle complex driving behavior of correlated context random variables, the dynamic Bayesian networks are used for the reasoning and implementation of the impact of contextual interaction. The simulation confirms the proposed approach improves the tracking accuracy and gets a better prediction of the uncertainties motion.