Pixel-Based Classification of Olive Trees Using Multi-Spectral UAV Images and Vegetation Indices

Olives are one of the world's most important harvested crops and the most studied fruit trees. In this context, studies on olive tree identification using state-of-the-art data acquisition methods are a prominent topic today. In this study, images obtained using an unmanned aerial vehicle equipped with a multi-spectral sensor were processed, and olive tree classification was performed using different methods. Using these methods, vegetation indices were generated using the spectral bands provided by the multispectral sensor, and three other data set combinations were created. The effects of NDVI, SAVI, DVI, and RVI indices on the classification were also investigated. Furthermore, the distinction between the olive tree class and other vegetation, artificial surfaces, and soil classes was also quantitatively examined in each dataset. Accordingly, the highest-performing classification was achieved with an overall accuracy of 91% using the Random Trees method with dataset set-2, which included multi-spectral bands.