Air-to-Ground Target Tracking in a GPS-Denied Environment using Optical Flow Estimation

This paper proposes a visual navigation system for an unmanned aerial vehicle to track a moving ground object in a GPS-denied environment. Image processing combines a target tracker which provides pixel coordinates of a ground target, and the estimation of optical flow around the detected target position. An extended Kalman filter is applied to estimate global position and velocity of the target as well as those of the own-ship aerial vehicle by fusing the image processing outputs with onboard inertial sensor measurements. The image processor and the estimation filter are tested on onboard camera images and vehicle state data that are synchronically recorded in actual flights.