Effects of changing canopy directional reflectance on feature selection

A Monte Carlo model was used to predict the mean apparent directional reflectance of a simulated plant canopy and the covariance for seven wavelength channels in the visible portion of the spectrum. The non-Lambertian spectral response from Bouteloua gracilis canopies possessing two plant cover densities was simulated for two solar positions. The calculated spectral signatures as a function of look angle were then analyzed using the divergence criteria to select the best two wavelength channels for discrimination. These calculations indicate that different combinations of wavelength channels are appropriate for various sensor look angles, that target signatures have greater statistical separation for some scan angles than others, and that these effects are time varying.