A device is arranged to carry out a method for identifying traffic-relevant information in a moving vehicle, wherein image data of a visual sensor and map data of a navigation system are each pre-evaluated for the identification and the results of the pre-evaluation are combined and interpreted. The image and map data is scanned for traffic-relevant information and compacted to relevant image and map data in a first and second scene interpretations, and the relevant image and map data is supplied to a rule engine and to a state machine for interpreting the image and map data. The rule engine evaluates the image and map data and transmits an action to be executed to the state machine which carries out a transition in its state space including pre-defined states and outputs information which is assigned to the state or to the transition.