ASSESSMENT OF VEGETATION INDICES FOR THE DETERMINATION OF AGRICULTURAL CROP TYPES

Abstract

With the increase of the world population, issues such as protecting natural resources and providing the sustainability of agricultural activities have gained importance. For the last four decades, observation of agricultural areas and generation of crop type maps are possible with remote sensing technologies. A commonly known method for determining the crop type is to classify satellite images and generate thematic maps, which can be useful to produce data for agriculture policies. In this study, RapidEye satellite images of Menemen District of Izmir Province were classified using both pixel based and object based methods. The vegetation indices such as NDVI, GNDVI, and NDRE have been used for the purpose of increasing the classification accuracy. As a result, 5 different data sets have been classified as; original bands of RapidEye, NDVI, GNDVI, NDRE, and the combination of all available bands together. Accuracy analyses have been performed by creating error matrices and kappa statistics. The results showed that GNDVI classified with the object based algorithm has the highest accuracy of 89.7744%. It is followed by the original bands together with all the indices, using both pixel based and object based classifications with the accuracies of 89.4486 and 89.3985%, respectively.