Land cover and crop classification using multitemporal sentinel-2 images based on crops phenological cycle

Italy is considered one of the developed country in the field of agriculture in Europe. For many reasons, reliable classification of crops and mapping plays an important role in Precision Agriculture (PA). Increasing availability, improving spatial resolution and high revisit time of sentinel-2 satellite become more useful and play an important role in analyses for land cover, crop classification and other remote sensing applications. Most of the Crops with similar spectral characteristics can be distinguished by accumulating spectral information of different phenological stages. In literature, many solutions have been proposed to classify crops using multitemporal images acquired from various satellites equipped with multispectral imagery sensors. However, features and images selection from multispectral, multitemporal images still needs improvement. In this paper, crops phenological cycles are investigated using temporal normalized difference vegetation index (NDVI) patterns and major crop phenological information are used to select best multitemporal images for the input of random forest (RF) classifier to classify land cover and crops.