Compensation for Azimuth Angle or Scale Effect on Building Extraction in Urban Using SAR Scales of Polarization and Coherence

In the building extraction of urban, polarization and coherence of synthetic aperture radar (SAR) data as multiple scales were used to compensate for the azimuth angle variation that is considered as the azimuth scale effect. As scales varied from single, dual, and multiple polarizations, and to polarimetry, the delineation of buildings with azimuth angles of 90° or slightly away from 90° was satisfactory. As the complexity of building patterns increased, the use of polarized data and polarimetric data even after the deorientation procedure became ineffective. Coupled with coherence data, we introduced a model to elevate the double-bounced scattering but to suppress the volumetric scattering from buildings of variable azimuth angles. Evaluated by polarimetric data from trihedral and dihedral corner reflectors (CRs) at Goldstone, CA, USA, the model increased the double-bounced scattering but decreased the volumetric scattering only for the dihedral CRs with an azimuth angle at 45°. The model had little effect on the trihedral CRs and the dihedral CRs with an azimuth angle of 90°. Results were satisfactory in the delineation of buildings with variable azimuth angles from vegetation using L-band E-SAR data from Oberpfaffenhofen, Germany. Thus, with multiple scales expressed as the single-, dual-, and multiple-polarization data, polarimetric data, and volumetric scattering model coupled with coherence information, the azimuth scale effect was sequentially removed.