The problem of classification and justification of the methodology for selecting classes of objects subject to recognition in a multispectral space image

Abstract

In the article, the specific task of Remote Sensing is being investigated to classify recognizable classes from objects consisting of 12 different types of vegetation soil features. The stages of solving this problem were selected and justified and the territorial areas where the test areas were formed for the development and testing of the classifier were determined. Based on information obtained from satellite and IKONOS, tables are provided that provide quantitative assessments for various class combinations of vegetation objects in the images formed from the satellite image of the studied area. The tables reflect statistical characteristics allow for preliminary assessments of the representativeness of the clusters. The article presents combinations of histograms for various variants of combinations of classified vegetation classes and provides tables of unsuitability using decision rules based on statistical characteristics.