|
The atmospheric Greenhouse gas Monitoring Instrument (GMI) on the Gaofen-5 (GF-5) satellite measures the reflected light from the sun in the near-infrared band and retrieves the concentration of major greenhouse gases such as CO2 and CH4 in the atmosphere. Because GMI does not image ground and lacks the coaxial visible light load, the captured spatial heterodyne spectral image cannot be registered with the control points like traditional remote sensing images. It can only rely on the calibrated location system on the ground for positioning, and there may be deviations in-orbit operation. This paper uses multi-source remote sensing data to design a location registration algorithm for GMI after it is in orbit. First, a region with a large continuous coastline is selected as the registration site, and the supervised learning algorithm is used to extract coastline data with high precision from the multi-spectral image of the China-Brazil Earth Resources Satellite4 (CBERS-04) as the spatial reference. Secondly, GMI O2 band data and DPC polarization multi-spectral data are used for joint cloud detection to filter out the pixels contaminated by the cloud. Finally, according to the spectral change characteristics near the coastline of the GMI route, the observation point located on the coastline is judged, the coordinates of the point are compared with the actual coastline point coordinates, and the coordinate offset is calculated to fit the position correction parameters. The results of experiments using on-orbit data show that the algorithm in this paper can better correct the GMI location error, provide a reference for the design of the correction algorithm of the same type of instrument, and propose improvements to the design of the next generation of related instruments.
|
