A Novel Drogue Pose Estimation Method for Autonomous Aerial Refueling Based on Monocular Vision Sensor

Drogue pose estimation is one of the key problems of autonomous aerial refueling (AAR) during the docking phase. A novel pose estimation method is proposed to remedy the failure of current methods due to the deformation of the flexible drogue. First, the key to flexible drogue pose estimation is suggested, and the concept of space-invariant features is extended. For the construction of 3L features, a preprocessing method is then proposed. Next, the problem of skeleton pose estimation is transformed into a fourth-order polynomial by constructing virtual points and presenting a point feature-based method. Subsequently, a matching method and the projection invariants of the skeleton are proposed. After that, the correspondences of reprojection points are constructed using the matched 3L features. An objective function fusing circle and line features and their derivatives are developed for optimizing the drogue pose base on the correspondences. Ultimately, the key techniques are verified in the synthetic and real data.