Discrimination and classification of vegetation species with multi-spectral camera by using unmanned aerial vehicles

Nowadays, discrimination of vegetation species on earth is used extensively in remote sensing studies. Remote sensing studies using satellite imagery are inadequate in studies where high accuracy of spatial and temporal resolutions are required due to the fact that satellites are not always able to obtain images from the same location and due to ground sampling distance. Therefore, in our study, UAV (Unmanned Aerial Vehicle) technology, which has been increasing and becoming widespread recently, has been preferred because of its flexible mobility. In the last decade, with the development of UAV technology, its use has become widespread in many civil application areas as well as in agricultural activities. It is used in many agricultural activities such as disinfestation, land use detection, drought damage detection, plant health monitoring. Multispectral cameras mounted in UAVs enable the identification of land use and vegetation types. In this study, multispectral images were obtained from the field where clover and soybean plants were planted in order to identify the plant species on the map. Images were processed in various software and classified with three different algorithms. The accuracy of these classification studies was examined. The same training sites and random sampling points were used in the classifications. As a result of the classifications, accuracy was calculated as 87% for Maximum Likelihood and 81% for Support Vector Machine. According to this statistical information, the best-performing classification algorithm for our study was observed to be is Maximum Likelihood.