Polarimetric roll-invariant three-component scattering model and application to terrain classification

Abstract

Inspired from the Freeman-Durden three-component model, a roll-invariant three-component scattering decomposition model is proposed for polarimetric SAR data in this paper. Three points of modification are included in the model. Firstly, the deorientation procedure is introduced into the scheme to make the decomposition results roll invariant. Secondly, a modified volume scattering covariance matrix is proposed to better match the deorientation volume scattering. Lastly, the symmetry test is introduced to measure the reflection symmetry of the scatterer. Then the roll-invariant model is combined with the maximum likelihood classifier to apply to the terrain and land-use classification.