ESA GNC Conference Papers Repository
Optical Flow based GEO Satellite Attitude Estimation from Payload Image Data
Geostationary Earth observation weather satellites acquire multi-spectral images during operational mode using a line scanner. Satellite rotational and translational motion during the image acquisition phase cause geometrical image distortions. Currently this problem is solved by the principle of landmark tracking using ground control points (GCP). It uses feature recognition (e.g. edge detection) of shorelines , , river or roads  or terrestrial landmarks  and matches the detected features with corresponding reference features taken from a database. Significant contributors to the GCP based image navigation performances are landmark changing over time, image distortions (blur, lightning, and cloud coverage) and imprecise knowledge of the landmarks absolute position. To overcome the limitations the current paper proposes an alternative attitude estimation approach using the whole image information of an imaging payload. The well known egomotion determination via optical flow (also image flow) ,  processing is adapted to scan images. This method offers several benefits against landmark tracking: it allows a robust and high accuracy determination of the satellite attitude as well as a geometrical correction of the payload scan images and it offers variants with different level of a priori information. The navigation methods proposed in this paper have been already implemented for on-ground operation, but on-board operation can be envisaged as a future goal.