Visual detection of air-to-air refueling drogue

Kurzfassung

We present an approach for the automatic visual detection and position estimation of an air-to-air refueling drogue for unmanned aircraft. The drogue, which is dragged by an ahead flying tanker, is recorded by a passive color camera. Based on the flashy color of the drogue, the image is segmented in probable drogues and background. The drogues prediction is specified by a circle shape identification, to determine the drogue position in 3D camera frame also if parts of the drogue are covered. Finally, the drogue is tracked and classified over several images, to sort out one-time identifications. The presented approach is validated with a simulation setup that bases on real flight test experiences. The validation shows that the presented method generally enables a detection of the air-to-air refueling drogue. The results are discussed with respect to the effect of covering the drogue by tanker or image frame and the influence of the blooming effect during the distance estimation.