GRSG 34th Conference 2023

Title: Variation in multiple outdoor Specim SWIR images acquired over a Single Day

Author: Tingxuan Jiang

Geologic remote sensing studies encounter variations in recorded data due to changing data acquisition environments, impacting overall brightness and mineral absorption features. These changes can reduce calibration accuracy and lead to misclassification. To mitigate these effects, understanding the systematic influences of solar movement and surface moisture is crucial.

This paper presents the effects of changing solar position and rock surface moisture conditions using outdoor hyperspectral measurements of rock gravel. The gravel is laid out in four pyramid-shaped piles of rock gravel outdoors. Two piles are made of Carrara marble split (one dry, one wet) and two piles are made of dark basalt split (one dry, one wet).

We conduct the measurements using a downward looking Specim SWIR camera mounted on an Artscanner rack. A Spectralon white reference bar at the pile’s level facilitates conversion of the camera’s digital numbers to reflectance. Before each measurement, we add water to the wet piles for maintaining moistened conditions. Measurements are acquired from ~10 am to ~6 pm at half-hour intervals. The data analysis starts from understanding the variation of measurements on the white reference over one day, because this is the most uniform and supposedly flat part of the spectrum. Subsequently, we focus on explaining how any white reference variations affect the rock spectra. Following steps involve the analysis of how solar position and rock surface moisture affect classification.

We plan to present three main results in this paper: 1) If and how solar position affects rock spectra and classification. 2) How wet rock surface varies the rock spectra and classification. 3) We will show if the classification robustness on the dataset is different for classifiers with different working principles, and explain the different robustness if there is.

Geologic remote sensing studies encounter variations in recorded data due to changing data acquisition environments, impacting overall brightness and mineral absorption features. These changes can reduce calibration accuracy and lead to misclassification. To mitigate these effects, understanding the systematic influences of solar movement and surface moisture is crucial.

This paper presents the effects of changing solar position and rock surface moisture conditions using outdoor hyperspectral measurements of rock gravel. The gravel is laid out in four pyramid-shaped piles of rock gravel outdoors. Two piles are made of Carrara marble split (one dry, one wet) and two piles are made of dark basalt split (one dry, one wet).

We conduct the measurements using a downward looking Specim SWIR camera mounted on an Artscanner rack. A Spectralon white reference bar at the pile’s level facilitates conversion of the camera’s digital numbers to reflectance. Before each measurement, we add water to the wet piles for maintaining moistened conditions. Measurements are acquired from ~10 am to ~6 pm at half-hour intervals. The data analysis starts from understanding the variation of measurements on the white reference over one day, because this is the most uniform and supposedly flat part of the spectrum. Subsequently, we focus on explaining how any white reference variations affect the rock spectra. Following steps involve the analysis of how solar position and rock surface moisture affect classification.

We plan to present three main results in this paper: 1) If and how solar position affects rock spectra and classification. 2) How wet rock surface varies the rock spectra and classification. 3) We will show if the classification robustness on the dataset is different for classifiers with different working principles, and explain the different robustness if there is.