The influence of changing moisture content on laboratory acquired spectral feature parameters and mineral classification

Abstract

Moisture content in soils, regolith, and rocks influences hyperspectral measurements and subsequent mineral classification. It is important to understand the systematic effects of changing moisture content on spectra and on the classification accuracy of different classification techniques. In this study, we measure reflectance spectra from six pure powder samples with moisture content varying from complete saturation to oven-dried. We analyse the change in the centre position, depth, and area of diagnostic spectral absorption features with changing moisture content and classify the spectra using the full wavelength range (350–2500 nm) and a subset (2100–2360 nm) with three classifiers that are sensitive to different parameters: (1) the spectral angle mapper that is only sensitive to colour changes; (2) the Euclidean distance classifier that is sensitive to both colour and brightness and (3) an expert classification system that uses only the centre position of diagnostic features. We find that, while the moisture content is below a specific threshold, the centre position of spectral features remains consistent as moisture content increases. However, once this threshold is exceeded, water absorption features become dominant over mineral absorption features, changing the absorption feature parameters dramatically. Moisture content affects subsequent mineral classification. Apart from the combination of materials observed, the chosen classifier type and wavelength range used are of great influence. We find that the centre position of spectral features is the least affected absorption feature parameter. The application of continuum removal in a narrow selected wavelength range helps mitigate the effect of changing moisture content.

Keywords

• Moisture content
• Hyperspectral
• Mineral
• Absorption features
• Classification