Airborne Traffic Flow Data and Traffic Management

Imagery collected from airborne platforms has long been used to document the evolution of traffic congestion over time and across extended areas. The authors have documented such published research dating all the way back to 1927, and these activities continued in various periods of activity up through the late 1980s. Beginning in the late 1990s, many important ideas, new and old, on the use of airborne imagery for traffic analysis were investigated and tested in the field by a number of research teams around the world. Through detailed image processing techniques, this imagery can be used to automatically determine traffic flow measures. Prototype software tools have been developed to automatically estimate queue lengths at intersections, to estimate vehicle speeds, and to estimate vehicle flows and densities. Similar research has integrated traffic flow data from airborne imagery into formal data collection programs, where the data from the imagery is fused with ground detector data to enhance the estimates and forecasts of traffic flows. Some of these research teams have also “architectured” approaches to automatically, in real-time, georeference images from remote cameras for managing traffic. This is done by integrating the imagery with information on the height, location, and orientation of the camera. Using these camera data, in combination with a geographic representation (latitude–longitude) of the area to be studied, lead to an explicit way to georeference the road and vehicle locations. Absolute values of vehicle positions, speeds, accelerations, decelerations, and lane changes can be determined. There have been a wide variety of experiments in this area; these tests have extracted individual vehicle trajectories from digital video imagery. Uses of the obtained airborne traffic data are many: (a) off-line analyses can be used for transportation planning purposes, (b) online, real-time analyses can be used for traffic management, especially in areas where ground sensors are not available and for large-scale incidents, and (c) one can develop data sets of vehicle trajectories for use in the calibration–validation of micro-, meso-, and macroscopic traffic flow models.