Models of Representation of Environmental Objects Based on Hyperspectral Survey Data

The work is devoted to the development of mathematical models for the representation and classification of various environmental objects based on hyperspectral imaging data. The work is aimed at selecting methods for representing and visualizing brightness features of environmental objects in hyperspectral images and developing a method for reducing the dimensionality of the spectral vector for classifying environmental objects. The paper describes the features of hyperspectral data. The concepts of the spectral image of a pixel and an object and the concept of the spectral model of the depicted environmental objects are presented. The methods of representation and visualization of brightness features of environmental objects identified in hyperspectral images are considered. The following types of spectral models are described: statistical, signature, and combinatorial. A method for reducing the dimensionality of a spectral model when processing large volumes of data is described, combining many computational blocks (heuristic search for optimal channel combinations, hybridization of simulated annealing and ray search methods, etc.). An approach to storing spectral models for classifying environmental objects is described. An experiment is conducted to calculate and compare spectral models of environmental objects, as well as to determine optimal combinations of spectral channels for describing environmental objects.