Accuracy improvement of geometric correction for CHRIS data

Abstract

This paper deals with a new type of multi-angle remotely sensed data—CHRIS (the Compact High Resolution Imaging Spectrometer), by using rational function models (RFM) and rigorous sensor models (RSM). For ortho-rectifying an image set, a rigorous sensor model-Toutin’s model was employed and a set of reported parameters including across track angle, along track angle, IFOV, altitude, period, eccentricity and orbit inclination were input, then, the orbit calculation was started and the track information was given to the raw data. The images were ortho-rectified with geocoded ASTER images and digital elevation (DEM) data. Results showed that with 16 ground control points (GCPs), the correction accuracy decreased with view zenith angle, and the RMSE value increased to be over one pixel at 36 degree off-nadir. When the GCPs were approximately chosen as in Toutin’s model, a RFM with three coefficients produced the same accuracy trend versus view zenith angle while the RMSEs for all angles were improved and within about one pixel.