INVESTIGATION OF THE EFFECTIVENESS OF MULTI- AND HYPERSPECTRAL FEATURE SYSTEMS IN THE CLASSIFICATION OF NATURAL AND ANTHROPOGENIC OBJECTS
V. A. Aksenov, S. I. Orlov, O. I. Potaturkin, S. B. Uzilov
Institute of Automation and Electrometry, Siberian Branch, Russian Academy of Sciences, Novosibirsk, Russia
Keywords: classification accuracy, natural and anthropogenic objects, multispectral images, spectral channels, feature systems, image processing
Abstract
It is proposed to use spectral channels with a width of 30 nm as feature systems for classifying natural and anthropogenic objects in the field of observation. The effectiveness of using algorithms that select and process a small number of narrow-band multispectral images of real scenes formed in the most informative intervals of the visible and near-infrared ranges has been experimentally demonstrated. A system of three features is defined that ensures high accuracy in classifying objects in real scenes. The possibility of switching from systems of three features to systems of two features in the classification of combined classes of test regions of interest is shown.
|
