Adaptive region of interest estimation for aerial surveillance video

In this paper, we propose an adaptive region of interest (ROI) estimation for aerial surveillance video. The adaptive ROI estimation algorithm first analyzes how the camera is used, such as pan, tilt and zoom control parameters, from the ego-motion computed from the aerial surveillance video. Using a frame based camera operator attention model, we can not only estimate the ROI of the camera operator at the frame level, but also estimate the ROI at the sequence level. Experimental results using video captured by predator unmanned aerial vehicle (UAV) show that the proposed adaptive ROI estimation algorithm is both effective and efficient.