GEO Satellite Image Navigation with Cloud Detection using Multispectral Payload Image Data
Earth observation from geostationary orbit requires extremely accurate pointing knowledge of the instrument. Due to misalignments, thermal distortion effects and uncertainties on the attitude and position, it is essential to use image information, to meet the stringent requirements. We developed a high accuracy attitude estimation system, which uses multispectral payload image data. The navigation system evaluates the shift of the image content and reconstructs the rotational motion of the satellite. The estimation results are filtered and fused with AOCS data. The geometrical image correction and registration of the satellite images is performed using the improved knowledge of the line of sight.
The reliability of image navigation depends fundamentally on the correct recognition of cloud areas in the image data. Clouds can cover the earth's surface and images can become incomparable as well as cloud motion can be interpreted as false satellite motion. Thus, a cloud detection algorithm has been developed, which is based solely on unregistered and non-calibrated image data. A multispectral analysis (MSA) algorithm has been implemented, which uses VIS and IR channels for cloud detection during day and night time. The paper presents a description of cloud detection algorithm, results of sensitivity analysis with respect to ground texture and lightning conditions and simulation results of the navigation performance under cloud conditions.