Investigation of the effects of vegetation indices derived from UAV-based RGB imagery on land cover classification accuracy

Abstract

The production of land use / land cover (LULC) maps using UAV images obtained by RGB cameras that offer high spatial resolution has recently increased. Vegetation indexes (VIs) are one of the important tools used to increase the limited spectral information of the UAV image in pixel-based classification. The aim of this study is to examine the effect of RGB-based VIs, called green leaf index (GLI), red-green-blue vegetation index (RGBVI) and triangular greenness index (TGI) which are frequently used in the literature, on the accuracy of thematic maps produced from UAV images. For this purpose, five different combinations comprising of RGB bands and VIs were formed. It was observed that the use of vegetation indices together with RGB bands increases the overall accuracy (OA) of the produced thematic maps in all cases. Additionally, the highest OA value was calculated from the thematic map produced using Dataset-5. The classification result of Dataset-4 consisting of RGB band and TGI, showed superior performance compared to Dataset-2 and Dataset-3, and a 0.1% difference was calculated between Dataset-5. Thus, this study has shown that the TGI index is more effective compared to GLI and RGBVI for thematic maps produced from a three-band UAV image.