GRSG Conference 2022: Orbit to Outcrop

Title: The Influence of Moisture Content on Mineral Classification of Hyperspectral Images

Author: Tingxuan Jiang

Abstract:

Remote sensing studies on minerals and/or rocks often involve a factor as moisture contents. Moisture content varies albedo of measurement objectives and affects spectral features of minerals and rocks. Therefore, for remote sensing studies, the changing moisture content can lead to lower calibration accuracy, noisier signal, and more misclassifications.

To reduce the effects, the systematic effects of changing moisture on mineral spectra need to be described and the principle of the effects needs to be explained. By referencing the systematic effects of moisture content, remote sensing studies can then separate spectral features caused by mineral differentiation out of moisture. The principle of the effect of changing moisture aids in developing models to predict and then correct the effect of moisture content.

In this paper, we present the systematic effect of changing moisture content on mineral spectra. We collected a lab-acquired dataset and then we will use the dataset to evaluate the effect on spectral features and classification results.

We measured seven pure mineral samples with changing moisture content in a dark room using an Analytical Spectral Device (ASD) TerraSpec Halo Mineral Identifier that is fixed in a solid position. The seven pure minerals are vivianite, gypsum, malachite, borax, calcite, kaolinite, and montmorillonite. We add ultra-pure water into the samples until a full saturation, and then frequently measure the mass and reflectance spectra of each sample while it is drying. We calculate sample moisture content by weight, therefore each sample is weighed before being measured.

Drying forms a crust of the samples, therefore we stir a sample before measurement to keep homogeneity. Every time a sample is measured, we rotate it four times and make three repeat measurements succeeding each rotation. When the drying of a sample stops, we oven-dry the sample at 105 ℃ for 24 hours (except for gypsum which is oven-dried at 75 ℃ to avoid dehydration) to make it completely dry. We weigh the mass of the oven-dried samples, and then use the mass of completely dried material to calculate the moisture content of the samples of each measurement.

The measurement results show two main effects of increasing moisture content on mineral spectra: 1) decreasing reflectance value; 2) decreasing absorption depth of mineral diagnostic features. Besides that, the effect of changing moisture content changes from mineral to mineral. For example, the diagnostic feature disappears when a mineral is fully saturated but the diagnostic feature of kaolinite is clear all the time. We are working on an explanation of the effects and then move to evaluate the effects of changing moisture content on mineral classification