Absolute velocity estimation of UAVs based on phase correlation and monocular vision in unknown GNSS-denied environments

This paper proposes a novel approach for absolute velocity estimation of unmanned aerial vehicles in unknown and unmapped GNSS-denied environments. The proposed method leverages the advantages of Fourier-based image phase correlation and utilizes off-the-shelf onboard sensors, including a downward-facing monocular camera, an inertial sensor, and a sonar. The non-matching tracking approach is particularly appealing, offering accurate estimation while remaining robust against frequency-dependent noise, significant intensity variations, and time-varying illumination disturbances. In the proposed method, the first step involves computing global pixel motion from consecutive images using phase correlation, which utilizes the shift property of the Fourier transform. This pixel motion estimation serves as the basis for creating a closed-loop solution for absolute velocity estimation. To further enhance accuracy, a Kalman filter is implemented to fuse all available data and provide a reliable velocity estimate. Validation of the proposed visual-inertial technique is conducted through simulation experiments using AirSim and real-world flight tests. The results demonstrate the practicality and effectiveness of the approach across a range of challenging scenarios. This paper proposes a Fourier-based image phase correlation method for absolute velocity estimation of unmanned aerial vehicles using off-the-shelf onboard sensors, including a downward-facing monocular camera, an inertial sensor, and a sonar in unknown and unmapped GPS-denied environments. The non-matching tracking approach is attractive and promising, with the advantages of accurate estimation, robustness against frequency-dependent noise, significant intensity variations, and time-varying illumination disturbances. image