An improvement of multiple-component scattering model with rotated covariance matrix for polarimetric SAR decomposition

Abstract

It has been validated that the cross-polarized scattering (HV) is caused not only by vegetation but also by rotated dihedrals, then the orientation angle of the rotated dihedrals can be obtained by minimizing the cross-polarized scattering component of covariance matrix. Therefore, this paper presents an improvement of multiple-component scattering model of rotated covariance matrix for Polarimetric SAR decomposition in order to detect the rotated buildings. Based on this model, the oriented buildings can be distinguished from the volume scattering mechanism of forest. Comparisons of the multiple-component decompositions with and without rotation of the covariance matrix are conducted using ESAR L-band Polarimetric SAR data of the Oberpfaffenhofen Test Site Area. Experimental results indicate that the improved decomposition model by implementing a rotation of the covariance matrix can recognize the oriented buildings from volume scattering and achieve a better decomposition result and further more accurate interpretation.