Airborne Monocular Vision Guidance Method for Autonomous Landing

Abstract

Vision guidance has shown increasingly promising capacity in autonomous landing, since it has the merits of low-cost and electromagnetic resistance. In this work, we propose an airborne monocular vision guidance method using region and structured line features. In the approach phase, a region-based pose tracker is adopted to track the pose within the consecutive frames using the contour and color information of the carrier. In the proximity phase, when the structural features of the carrier become distinguishable, the lines forming a particular configuration are detected. Then, a similarity measurement criterion based on direction and distance constraints is exploited to perform line alignment between 3D lines and the detected 2D ones. Once the 2D-3D correspondences are identified, the rotation and translation are successively computed using the structural line constraints. Finally, the optimizer of the pose tracker is utilized to refine the pose for better accuracy. Experiments demonstrate that the proposed method achieve high pose estimation accuracy and real-time efficiency, which is suitable for guidance of autonomous landing.