Abstract

Persistent surveillance of large geographic areas fromunmanned aerial vehicles allows us to learn much about thedaily activities in the region of interest. Nearly all of the aproaches addressing tracking in this imagery are detection-based and rely on background subtraction or frame differ-encing to provide detections. This, however, makes it diffi-cult to track targets once they slow down or stop, which isnot acceptable for persistent tracking, our goal. We present a multiple target tracking approach that does not exclusively rely on background subtraction and is bet-ter able to track targets through stops. It accomplishes this by effectively running two trackers in parallel: one based on detections from background subtraction providing target initialization and reacquisition, and one based on a target state regressor providing frame to frame tracking. We evaluated the proposed approach on a long sequence from a wide area aerial imagery dataset, and the results show improved object detection rates and ID-switch rates with limited increases in false alarms compared to the competition.