Singular value decomposition in multispectral radiometry

Abstract

Abstract A method is defined for the estimation of the strength of signal present in spectral data sets which are dominated by background and noise effects. The method is particularly well suited to data in cases where a considerable body of observations is available for calibration purposes. The spectral characteristics of the required signal as it affects the data, the 'key vector', is determined by Singular Value Decomposition and optimised using the known constituent data. A simulation is performed to illustrate the determination of the key vector and its use to estimate constituent content on subsequent data sets.