Новітній підхід до застосування нормалізованого диференційного вегетаційного індексу для класифікування аерокосмічних зображень за наявності неповних та неточних даних

Abstract

Solution of different scientific, ecological and agricultural tasks with the use of aerospace im-ages comprises a procedure of image classification. Classification is one of the most important procedures. Nowadays many supervised and unsupervised classification methods are applied in remote sensing. The most accurate results are obtained through the use of supervised classifica-tion methods. In this paper, there are proposed some new approaches to image classification which are based on supervised classification methods and Normalized Difference Vegetation Index (NDVI). Different values of NDVI are noted to correspond to different classes of objects, such as soil, water, roads, sand, green vegetation, oil spills. Application of Vegetation Index is the first step of classification. Using NDVI, it is possible to select special necessary classes. Af-ter the application of NDVI, such classification methods as the parallelepiped method, Demp-ster’s rule, and Inagaki’s combination rule can be used. The current work describes the main advantages of these classification methods. It has been noted that the use of the parallelepiped method allows easy and quick processing of data. The paper also shows that Dempster’s combi-nation rule and Inagaki’s combination rule can deal with inaccurate and incomplete data from different spectral bands. Moreover, these methods can process conflicting information. Demp-ster-Shafer theory has the advantage of high accuracy and simple calculations. In the paper, there is also considered a numerical example where NDVI and Inagaki’s combination rule has been used for detection and mapping of oil spills. Application of Vegetation Index and such su-pervised classification rules as the parallelepiped method, Dempster’s rule, Inagaki’s combina-tion rule can be applied in ecological monitoring, mapping of petroleum spills, and solving ag-ricultural tasks.