The problem of remote sensing of soil and vegetation cover using airborne hyperspectral cameras is considered. The modern state of development of airspace measuring instruments with high spectral resolution is analyzed. The problems and solution methods of the thematic processing of hyperspectral images are discussed. Applications of the optimal Bayesian classification for the recognition of the forest stand species are investigated. The method of the effective reduction of the dimensionality of the feature space is proposed. The classification accuracy for the selected test areas is estimated.