IRAL: Robust and versatile UAV localization using infrared vision and altitude sensor fusion

Supplementing or replacing the Global Satellite Navigation System (GNSS) for robust UAV localization remains a challenge. In this work, we propose an infrared vision and altitude sensor fusion method called IRAL, which mainly includes on-board near-infrared (NIR) beacons, off-board narrowband-pass vision sensors with the same wavelength as the beacon and strapdown high precision altitude sensors. The beacon with a high signal-to-noise ratio as a cooperative target provides robust features, thereby facilitating beacon recognition through the designed gradient-based sequential frame template matching (GSFTM) algorithm. The proposed method measures the altitude difference between the UAV and the vision sensor through the altitude sensor to accomplish depth estimation. After obtaining the beacon's pixel coordinates and depth, combined with the intrinsics and extrinsics of the vision sensor, the observation equation can be set up to solve the UAV's spatial position. Real-world experiments under various scenarios demonstrate that the proposed method stably achieves high accuracy.