Processing UAV Based RGB Data to Identify Land Cover with Focus on Small Water Body: Comparison of Methods

Abstract

Multispectral satellite imagery is frequently used to land cover identification. The higher spatial and temporal resolution may be needed to observe small features located on the Earth surface. It leads to the necessity of lower elevation data collected on demand. Airborne or UAV imagery meets these requirements. A middle- or low-class unmanned aerial vehicle (UAV) carrying a common RGB camera can be used to collect data suitable for identification of key land cover types. UAV is used as a source of data with a very high spatial resolution collected on demand. This solution is suitable for smaller areas and it is limited by legal regulations and weather conditions. The paper is focused on land cover identification near a small water body by various methods: Unsupervised classification, supervised classification, and vegetation (spectral RGB) indices. In the end, the used methods are compared.