DIRS Laboratory 76-3215
October 18, 2018 at 1:00am
GREG BADURA
Ph.D. Thesis Defense
Abstract: 

A major focus of the Goniometer of Rochester Institute of Technology (GRIT) Laboratory is using Hapke’s photometric model to retrieve geophysical parameters of sediment surfaces from bidirectional reflectance measurements. An important parameter of the model is the mean slope angle roughness metric.  A series of laboratory experiments was performed to isolate the mean slope angle parameter and correlate the parameter to observed spectral phenomena. In the first experiment, BRF and surface digital elevation measurements were performed on dry clay sediments of varying roughness. The Hapke mean slope angle parameter was derived for each sample. We found that spectral variability, especially near spectral absorption features correlates strongly with quantified measures of surface roughness. This suggests that roughness parameters used in some radiative transfer models, such as the Hapke model, might be directly determined from the spectrum itself. In the second experiment, assumptions made by Hapke in deriving the photometric roughness correction are tested by generating sand samples of constant sample density and grain size distribution, but varying roughness. The type of roughness was also classified into two different cases: “wave-like” and “grid-like.” The “grid-like” roughness parameter meets the criterion outlined by Hapke in his correction factor, while the “wave-like” roughness parameter does not. By experimentally forward propagating Hapke’s roughness correction factor for the “grid-like” roughness samples, we find that the Shadowing function potentially does not account for centimeter scale roughness accurately. By examining the bidirectional reflectance measurements of the “wave-like” roughness at different orientations to the principal plane, we found evidence that the multiple scattering term should be incorporated into future correction factors for surface roughness. A third experiment that outlines a processing chain for deriving structural parameters of marshgrass vegetation using similar computer vision and data science techniques will also be discussed.