Vision-based dual-station airplane flight pose measurement via dense correspondences

Pose measurement is a crucial task in airplane flight testing. Among the existing methods, off-board (ground-based) visual approaches stand out for their high capability, flexibility, and non-contact operation. However, single-station single-view systems suffer from ill-posed problems in distance estimation, while single-station multi-view systems are constrained by short baselines due to their size, both resulting in poor position measurement accuracy. To address these problems, this paper proposes a novel dual-station pose measurement method. First, the two single stations independently perform coarse pose estimation and dense 2D-3D correspondence extraction, reducing the search space and enhancing robustness. Next, triangulation and attitude averaging are employed to fuse the pose estimates from the two stations. Finally, global optimization of the airplane poses is conducted based on the dense correspondences obtained, with dual-station structural parameters included in the optimization to mitigate their impact. The proposed method was evaluated through numerical simulations and real flight tests, demonstrating 0.45^∘ attitude measurement accuracy, 1.97 m positioning accuracy, and a computation speed of 20 Hz within a 1.6 km intersection area.