On the capability of different SAR polarization combinations for agricultural monitoring

This paper examines the capability of different SAR polarization combinations for agricultural monitoring. For this purpose, a time series dataset of five quad-polarized images acquired by RADARSAT-2 (C-Band) is used. The different SAR polarization combinations are generated by splitting each input dataset in two additional dual-polarization combinations synthetically. Polarimetric decomposition is realized by a new Kennaugh matrix like decomposition, while the mandatory speckle filtering is performed by a pyramidal multi-looking approach. Thus, the data is normalized in order to fulfill the requirement of a normal distribution for the subsequent maximum likelihood classification. Concluding, the accuracy assessment provides a measure for the questioned classification capability of dual-polarized images in comparison to the quad-polarized data.