Multi-temporal indices derived from time series of Sentinel-1 images as a phenological description of plants growing for crop classification

Agricultural land cover is characterized by fast changes within time. The phenological dynamic of plants can deliver crucial information for crop classifications. To enhance a crop classification multi-temporal indices are proposed. They are calculated based on the time series of coherence matrices and parameters of H/alpha decomposition derived from dual polarimetric synthetic aperture radar images (Sentinel-1). The study shows that the use of multi-temporal indices increases the accuracy of crop classification by 9% in comparison to classification based only on the time series of coherence matrices.