UAS-~based Object Tracking via Deep Learning
Abstract
Tracking is the task of identifying an object of interest and detect its position over time, and has numerous applications like surveillance, security and traffic control. In present times, unmanned aerial vehicles (UAV) have been more and more common which provides us with a new and less explored domain, with an ideal vantage point for surveillance and monitoring applications.. Aerial tracking is a particularly challenging task as it introduces new environmental variables such as rapid motion in 3D space. We propose a new deep learned tracker architecture catered to aerial tracking.
First, a study of six state-of-the-art deep learned trackers has been performed using the Visual Object Tracking benchmark. This study determined the weaknesses of said trackers in front of a long-term aerial tracking task. Mainly, severe motion, target disappearance and high degree of appearance change were the principal causes for drift or loss of track.
Siamese correlation filter based tracker to perform identification and target matching across subsequent frames. In addition, a multi-scale object detector has been implemented to improve identification accuracy and template update. The object detector goal is to output a score map that will validate or penalize the tracker's correlation map, and improve robustness against drift, scale change and occlusion.