USE OF CROSS-POL MULTI-TEMPORAL SAR DATA FOR IMAGE SEGMENTATION

The contribution to the image segmentation of multi-temporal SAR data acquired with VV - VH polarization is analyzed. The backscattering intensity, the short time interferometric coherence and the decorrelation time of scatterers have been used to separate different ground backscattering behavior both in VV and VH polarization. Then, these six parameters have been projected into a domain with reduced dimensionality by means of a Covariance Matrix Analysis. The resulting covariance values show that the main contribution to the image segmentation from the VH polarization comes when the short time interferometric coherence is different from that of the VV polarization. Preliminary results are presented on an agricultural area. Moreover, the VV-VH complex coherence has been exploited to identify urban areas and permanent scatterers. It has been verified that the coherence phase in urban environments can be only 0 or pi accordingly to the theory of dihedral and trihedral backscattering.