Efficiency Assessment of Different Approaches to Crop Classification Based on Satellite and Ground Observations

A problem of crop plants classification for three regions of Ukraine with an area of 78.500 km(2) is considered. Classification is carried out using not a single satellite but a time series of satellite images. The used satellite data are characterized by different spatial resolution and temporal characteristics. By example of this problem we assessed the efficiency of different classification algorithms (neural networks, decision trees and support vector machines) for substantially different data levels (of training and testing samples) both for extremely large data sets and under the condition of their lack (absence).