Track based characterization of vehicle behavior

In many areas of the world, overhead video collected from aircraft and other airborne vehicles is an important component of surveillance and security operations. A typical use scenario might involve real-time monitoring of video streams by human analysts with the goal of identifying patterns of vehicle movement that might be considered suspicious or dangerous. In performing such a task, analysts often have access to low-level image processing support, most notably automated extraction of vehicle tracks. The goal of this research is to investigate the hypothesis that analysts would benefit from higher-level automated support that provides an assessment of vehicle behavior on the basis of an analysis of track data. Experimental results were obtained in the context of a simulated environment that involves a relatively simple task for video analysis. The results show that automated support for the identification of suspicious vehicle behaviors significantly improved successful vehicle identification.