Specifics of space image classification for forest identification of the Carpathian region

Abstract

The aim of this work is to study the effectiveness of the use of controlled classification to identify forest vegetation by high-resolution space images; identification of healthy vegetation, completely withered and damaged by drying conifers. Method. The study of the influence of the choice of the number of signatures for the controlled classification on the basis of the parametric rule of maximumprobability based on a high-resolution image obtained fromthe GeoEye1 remote sensing system. Results. The study is based on the analysis of statistical characteristics of the spectral brightness of pixels, which allows us to conclude about the priority of signatures of a particular size. The created classified images for two cases of the chosen sizes of signatures on test sites allow to estimate accuracy of the areas of the chosen classes. Scientific novelty and practical significance. The novelty of the obtained results is the study of the size of training samples for the controlled classification of space images by the method of maximum probability. The method of controlled classification according to the rule of maximum probability allows to identify various objects characteristic of the forest vegetation areas. Using the right selection of signatures and their location in the image, you can determine the type of forest objects, including categories of conifers: healthy, damaged and dry, which have complex spectral brightness. That is, the formation of training samples in the classification of forest objects with mixed spectral characteristics requires additional research