Crop Classification Based on Microwave Images Acquired from Envisat and Ers

Crop recognition on microwave satellite images was examined using ENVISAT and ERS images acquired during growing season of 2003. The necessary ground data were collected from a test site located in Poznań region, in the western part of Poland. The evaluation of crop discrimination was performed by analysing the appropriate cluster separability within the spectral space defined by the images. Classification was applied to mean values of the backscattering coefficient modified by the cosine of the microwave incidence angle and calculated for image segments representing homogeneous areas. Nonparametric supervised classifier, namely a multilayer perceptron neural network, was applied to the whole set of images, as well as to the selected subsets of images, in order to find the most suitable minimum of data for crop discrimination. The best results were achieved by applying HV and VV images collected at the end of April and in the beginning of June and July.