Classification of different types of snow using spectral and angular imaging

Sammanfattning: The current thesis work details a non-contact detection approach concerningclassification of snow with different physical properties such as grain size, densityand specific surface area (SSA). In this approach, reflected light from snowsurfaces is measured as a function of wavelength and viewing geometry. Essentiallya detector (either a near-infrared (NIR) camera or a spectrometer) and anillumination source are needed to measure the spectrally and angularly resolvedbidirectional reflectance from snow. Classification of snow types is performedbased on the absorption and scattering properties of a respective snow type. Itis furthermore known that snow properties can be modelled using a numericalsolver where the radiative transfer equation (RTE) for snow is solved and ascattering phase function is estimated by expanding into a series of Legendrecoefficients. It is therefore expected to be a connection between snow characteristicsand the Legendre coefficients of the scattering phase function.Results suggest that different snow types can be classified using two wavelengths(980 nm, 1310 nm) from the high reflectance region and one wavelength(1550 nm) from the high absorption region. It is also observed that thebidirectional reflectance for snow tends to increase in specular direction (antiilluminationdirection) as snow density increases. Results from the numericalmethod suggest that the first coefficient of the Legendre phase function is arelative estimate of the single scattering albedo rather than an absolute estimateand that the second coefficient estimates the anisotropy of a respectivesnow type. Investigations in this thesis suggest that the presented approachcan be used as a tool to classify different snow types in various applicationssuch as icing on wind turbine blades, winter roads maintenance and ski tracksmaintenance.v