Polarimetric calibration requirements on several classification schemes for land application of polarimetric synthetic aperture radar

Polarimetric calibration intercalibrating the polarisation channels of the radar system results in a better characterisation of the target and of the backscattering mechanisms. However, system distortion because of miscalibration may affect the classification performance to some extent. The study described in this article is an attempt at deriving the polarimetric calibration requirements on several algorithms for the classification of land features. In the monostatic backscattering case, the model representing the effects of the residual distortion of the calibrated polarimetric data is presented first, followed by analysing the effects of miscalibration on a few commonly used polarimetric parameters and several classification schemes. Experimental results show that polarimetric calibration requirements not only are strongly dependent on classification methods but also directly relate to the underlying physical scattering mechanisms of the observables. For each classification algorithm under study, a set of numbers summarising the polarimetric calibration goals concludes this study.