Tracking both pose and status of a traffic light via an Interacting Multiple Model filter

Either for driver assistance systems or autonomous vehicles, detecting traffic lights (status and pose) is required when Intelligent Transport Systems go downtown. As detection algorithms could still have some misclassification on the traffic light status, this paper proposes a solution to nearly avoid this problem. An Interacting Multiple Model filter is used to track both the position and the status of a traffic light through the time and to increase traffic light recognition performances for automation purpose.