Image Registration for Geostationary Orbit Satellite using Inverse Ortho Correction

Abstract

A geostationary Earth observation satellite with a middle-resolution optical system can observe a particular point at any time, but image registration and geometric correction are not easy because an oblique view is required when observing the Japan area from geostationary orbit. In this paper, we propose an image registration framework including image matching with other satellite images to perform image registration with high accuracy. In the proposed framework, the most distinctive feature is to perform an inverse-ortho correction before image matching to improve matching accuracy corresponding to oblique viewing from geostationary orbit. Additionally, the framework integrates a telescope alignment estimation method to accurately obtain line-of-sight direction as an alternative registration approach. This method entails fitting observed alignment values to a change model based on thermally induced errors in the telescope, ensuring precise alignment estimation. Our simulation results demonstrate the superiority of image matching using inverse-ortho correction compared to conventional standard-ortho correction. The proposed method achieves sub-pixel matching accuracy. Furthermore, we show that an alignment value can be estimated with an order of 10^-5 degrees by applying the proposed estimation method.