CALIBRATION ACCURACY OF AERIAL MULTISPECTRAL REFLECTANCE IMAGES AND ESTIMATION OF ERROR SOURCES

We have performed a systematic study of the calibration accuracy of multispectral aerial reflectance images. The study is part of the improvement of change detection for aerial and satellite multispectral images. It is based on (a) a multitemporal image data set from the multispectral airborne scanner Daedalus AADS 1268 and (b) simultaneously made ground based hyperspectral reflectance measurements. Reflectance images were obtained from the raw images by processing with the program package SENSAT-5, using standard urban atmospheres in the MODTRANcode. The systematic uncertainty of the resulting reflectance spectra is mainly due to the uncertainty in the atmospheric parameters. It can be reduced by fitting the Daedalus spectra to ground truth hyperspectral reflectance measurements of selected surfaces. The resulting uncertainty of the reflectance is estimated to be 11.0 % on average.