Drogue position and tracking with machine vision for autonomous air refueling based on EKF

For autonomous air refueling (AAR) of unmanned aircraft, the measurement of relative position between receiver aircraft and tanker aircraft is critical in the docking phase. This paper proposes a monocular vision-based relative position estimation method, which calculates the location coordinate by recognizing the beacons on the drogue with image processing. To increase the robustness for rejecting air disturbance and improve output frequency of drogue detection, the extended Kalman filter is applied to this system for estimating the states. The EKF can provides position estimation of drogue in camera frame if the image detection failed because of disturbance. And the region of interest (ROI) can be estimated according to the predicted drogue position in pixel frame, which reduce the time of image processing significantly. In order to evaluate the performance of this machine vision strategy for AAR based on EKF, a simulation validation platform is established. The real relative position can be obtained on this platform, the comparison between real values and estimated values indicates that the EKF has a high accuracy on drogue position estimation, meanwhile, the ROI can track the drogue smoothly.