Track based characterization of vehicle behavior

摘要

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.