Near-InfraRed and Visible (Vis-NIR)spectroscopy is a promising tool allowing to quantify soil properties. It showsthat information encoded in hyperspectral data can be useful after signalprocessing and model calibration steps, inorder to estimate various soil properties throughout appropriate statisticalmodels. However, one of the problems encountered in the case of hyperspectraldata is related to information redundancy between different spectral bands.This redundancy is at the origin of multi-collinearity in the explanatoryvariables leading to unstable regression coefficients (and, difficult tointerpret). Moreover, in hyperspectral spectrum, the information concerning thechemical specificity is spread over several wavelengths. Therefore, it is notwise to remove this redundancy because this removal affects both relevant andirrelevant hyperspectral information. In this study, the faced challenge is tooptimize the estimation of some soil properties by exploiting all the spectralrichness of the hyperspectral data by providing complementary rather thanredundant information. To this end, a new reliable approach based onhyperspectral data analysis and partial least squares regression is proposed.
展开▼
