AUTOMATIC TRAFFIC MONITORING WITH AN AIRBORNE WIDE-ANGLE DIGITAL CAMERA SYSTEM FOR ESTIMATION OF TRAVEL TIMES

Knowledge of accurate travel times between various origins and destinations is a valuable information for daily commuters as well as for security related organizations (BOS) during emergencies, disasters, or big events. In this paper, we present a method for automatic estimation of travel times based on image series acquired from the recently developed optical wide angle frame sensor system (3K = “3-Kopf”), which consists of three non-metric off-the-shelf cameras (Canon EOS 1Ds Mark II, 16 MPixel). For the calculation of overall travel times, we sum up averaged travel times derived from individual vehicle velocities to pass defined road segments. The vehicle velocities are derived from vehicle positions in two consecutive geocoded images by calculating its distance covered over time elapsed. In this context, we present an automatic image analysis method to derive vehicle positions and vehicle distances involving knowledge based road detection algorithm followed by vehicle detection and vehicle tracking algorithms. For road detection, we combine an edge detector based on Deriche filters with information from a road database. The extracted edges combined with the road database information have been used for road surface masking. Within these masked segments, we extract vehicle edges to obtain small vehicle shapes and we select those lying on the road. For the vehicle tracking, we consider the detected vehicle positions and the movement direction from the road database which leads to many possible matching pairs on consecutive images. To find correct vehicle pairs, a matching in the frequency domain (phase correlation) is used and those pairs with the highest correlation are accepted. For the validation of the proposed methods, a flight and ground truth campaign along a 16 km motorway segment in the south of Munich was conducted in September 2006 during rush hour.