Classification of the forest cover of Tver oblast using hyperspectral airborne imagery

Abstract

Recent research efforts have been focused on building a system of hyperspectral aerial sounding of forest vegetation on regional scales. The components of this system are developed using data obtained in the course of measurement campaigns in Tver forestry test sites. Hyperspectral airborne surveys are conducted using a Russian video spectrometer produced by the NPO Lepton company. The technique for recognizing ground-based objects is based on Bayesian classification principles with the feature space optimization. The choice of the most informative spectral channels is based on the step-up method. We propose an approach allowing the choice of channels to be more stable. We compare the classification of timber stands on the basis of hyperspectral imagery with ground-based data to demonstrate the consistency of the system developed.