Object recognition and correlation methods for traffic flow analysis

In comparison to its importance, the road transport system has a poorly developed information infrastructure. Therefore both the operators and users of this transport carrier have limited knowledge of the actual traffic conditions in the road network. If infrastructure is available for measuring traffic data, it is usually point-based and does not deliver parameters such as journey times of vehicles. Depending on the physical principle used, the sensors in many detectors provide more information on an individual vehicle than is generally used. However this information is mostly not suitable for identification of individual vehicles. The following approach uses pattern recognition and correlation methods to identify strings of vehicles in a similar manner to that used in communications engineering to identify bit strings for decoding messages. The run time of the message has its equivalent in the journey time of the vehicle flow. Some experimental results are described in the following, using sensor signals from a combined microwave and ultrasonic detector.