Improving quality of imaging spectroscopy data

Abstract

Imaging spectroscopy is moving into quantitative analysis of ecosystem parameters, which require high data quality. Thus, imaging spectrometers shall provide users with very accurate and low uncertainty measurements such that truthful products and reliable policies can be generated. However, the quality of imaging spectroscopy data, which can be interpreted as the distance between the measurement and the true value, depends on a series of disturbance factors that can be divided into instrument factors, environmental factors, and data processing factors. Those factors lead to data non-uniformities and inconsistencies that, if not properly identified, quantified, and corrected for, can compromise the quality of the scientific findings. This thesis investigates various techniques aimed to ensure the consistency of imaging spectroscopy data, namely in the spectral domain, throughout the data acquisition and specific data processing schemes, as for instance the calibration to radiances, with particular emphasis on instrument factors. The impact of data inconsistencies and non-uniformities on the quality of imaging spectroscopy data is first estimated. A scene-based technique for the characterization of keystone non-uniformity is then proposed. Moreover, a laboratory approach is established as the most reliable technique for the achievement of high accuracy calibration and characterization of imaging spectrometers. Last, an algorithm that identifies optimal sensor acquisition parameters for the retrieval of specific products in spectral regions of interest is presented. It has been concluded that laboratory calibration and characterization procedures offer a higher degree of fidelity with respect to scene-based methodologies when non-uniformities and calibration parameters have to be determined and implemented into correction schemes. A critical discussion of the main findings analyze advantages and drawbacks of the proposed techniques and suggests further improvements as well as future perspectives for the continuation of this work.