A PolSAR Decomposition Algorithm With A New Volumetric Scattering Model For Urban Target Delineation

In Polarimetric Synthetic Aperture Radar(PolSAR)data,backscattering from an urban target is highly influenced by the azimuth orientation angle(AOA)of the target.The AOA is defined as follows.When the wall of the building is parallel to the flight direction of the radar,the AOA is 0o.The wall and ground surface can form a dihedral corner reflector producing radar echo with strong double-bounced scattering component.When the wall and radar flight direction is not parallel,the AOA becomes non 0o.The echo from the interactions of the building and ground surface consists of cross-polarized scattering component.The cross-polarized backscattering is the strongest when the AOA is 45o.In addition,the cross-polarized and strong backscattering typically comes from tree canopies in forested environment.Thus,the cross-polarized backscattering is characterized as the volumetric scattering.In previous and developed model-based PolSAR decomposition algorithms,the urban targets are mainly delineated based on the existence of the strong double-bounced scattering,and vegetated surface the strong volumetric scattering.Due to the existence of strong cross-polarized backscattering from an ubrab target with non-zero AOA,the urban target can be incorrectly classified into the vegetation category.To reduce the misclassification,and to correct the mistake eventually for an urban target with non-zero AOA,one should decrease the volumetric scattering but increase the double-bounced scattering.The double-bounced scattering from the urban target should be larger than the volumetric scattering as well as than the single-bounced surface scattering.Thus,the urban target can be correctly identified.The PolSAR backscattering from an urban target has strong double-bounced scattering component when the AOA is near 0°,but has strong volumetric scattering when the AOA is about 45°.Based on the unique characteristics,the backscattering coefficients was first combined linearly,and a new correlation coefficient that describes the scattering mechanisms of urban targets with variable AOAs introduced.The value of the coefficient was high in urban area but low in non-urban area.The coefficient should quantify the scattering mechanism of the urban targets well regardless of their AOAs.Then,using the coefficient we developed a new volumetric scattering model that is integreated into an existing and well-known volumetric scattering model,and proposed a new four-component PolSAR decomposition algorithm.After analyses of US NASA/JPL AIRSAR data consisting of trihedral corner reflectors,dihedral corner reflectors leveled on the flat ground surface,and dihedral corner reflectors with 45° AOA,the algorithm had little effect on the single-bounced scattering from the trihedral corner reflectors and doublebounced scattering from the dihedral corner reflectors leveled on the flat ground surface.However,the volumetric scattering from the dihedral corner reflectors with 45° AOA decreased,and the double-bounced scattering elevated.Thus,the algorithm was able to categorize urban targets into an urban category regardless of their AOAs.To verify the effectiveness and applicability of the algorithm further,we applied it to three addtional PolSAR datasets acquired by the NASA/JPL UAVSAR of USA,E-SAR of Germany,and ALOS-PALSAR of Japan.Urban targets with variable AOAs were correctly classified into the urban category.Therefore,the algorithm has improved the identification rate in the urban target delineation.