Radiometric characterisation of a VNIR hyperspectral imaging system for accurate atmospheric correction

Abstract

The Institut Cartogàfic de Catalunya (ICC) regularly operates a Compact Airborne Spectral Imager (CASI) sensor. For this system an atmospheric correction algorithm was developed to simultaneously correct multiple overlapping images taken from different heights. First, the algorithm estimates the main atmospheric parameters with an inversion procedure using either radiometric ground measurements or image homologous areas plus a single ground measurement. Then, the code is applied to the images to obtain atmospherically corrected hyperspectral imagery. The algorithm was applied in the frame of ICC-Banyoles 2005 experiment (Spain) using multi-height imagery and field simultaneous reflectance measurements. In the validation step, the standard deviations obtained with both inversion methods were similar. In order to improve these results, the smiling effect (spectral shift) for the sensor is characterized by locating O2 absorption bands in the NIR for each CASI look direction. Additionally, a more accurate spectral sensitivity for each band has been calculated. These improvements are applied to EuroSDR-Banyoles 2008 experiment's (Spain) imagery. These results show a substantial improvement on the atmospheric correction at the absorption regions when compared to field reflectance measurements. This behaviour advises the inclusion of these developments in the inversion system.