Research On Scattering Characteristic Analysis And Classification Of PolSAR Targets Based On Azimuthal Symmetry

In PolSAR datasets,radar targets can be azimuthal symmetric or asymmetric targets for different physical features.The previous PolSAR decomposition algorithms were constructed based on the assumption of azimuthal symmetry.Therefore,there was always confusion about the scattering characteristics between the azimuthal symmetric and asymmetric radar targets,and there were also misclassification problems between these two types of radar targets according to the scattering characteristics.Lots of PolSAR decomposition algorithms tried to solve this problem by adding physical models.More models,and more execution complexity in the algorithm.Unfortunately,the results were not improved much.Therefore,for the development of the PolSAR decomposition algorithm,it is not only necessary to describe the scattering characteristics of the two types of radar targets correctly and classify them properly,but also to find a balance between the effectiveness and the execution complexity of the algorithm.Starting from the azimuthal symmetry characteristics of radar targets,a model-based PolSAR decomposition algorithm framework was developed.The algorithm enhanced the ability of the PolSAR decomposition algorithm to describe the scattering characteristics of radar targets,and improved the classification accuracy of radar targets in PolSAR data.The number of physical models was controlled,the execution complexity of the algorithm was reduced,and the practicability of the PolSAR decomposition algorithm was further enhanced.The main research contents of this dissertation were:(1)To make full use of the azimuthal symmetry information of radar targets in PolSAR data,this research carefully studied the relationship between the azimuthal symmetry and backscatter information of several typical radar targets in C-,L-,and P-band PolSAR datasets.An azimuthal symmetry factor(S_f)was proposed to describe the azimuthal symmetry of radar targets in PolSAR datasets.In PolSAR data,urban targets were azimuthal asymmetric radar targets having nothing to do with the radar wavelength.Airport runways/taxiways,ocean,and inland water surfaces,and grasslands were azimuthal symmetric radar targets,and the longer the wavelength,the better the azimuthal symmetry.The azimuthal symmetry of the forest targets was affected by the wavelength of the SAR.In C-and L-band PolSAR datasets,the forest targets were azimuthal symmetric radar targets,while in P-band PolSAR datasets,adversely influenced by the topography under the tree canopy,the forest targets will lose the azimuthal symmetry.Using 0.01 as the threshold,the delineation accuracy of azimuthally symmetric and asymmetric radar targets in C-and L-band PolSAR datasets can reach 92.7%.(2)To analyze the scattering characteristics of and classify the azimuthal asymmetric urban targets with large azimuth orientation angles in PolSAR data,this research used the diplane model to model the volume scattering characteristics of the urban targets.The polarization rotation method was used again to redistribute the volume scattering into the double scattering,which weakened the volume scattering and enhanced the double scattering of the urban targets simultaneously.Then the urban targets can be delineated from the forest targets with strong volume scattering.Finally,the diplane model was combined with the single,double,and volume scattering models to construct a four-component PolSAR decomposition algorithm.The algorithm has been successfully used to analyze the scattering characteristics of typical radar targets in multiple airborne and spaceborne C-,L-and P-band PolSAR datasets,such as urban and forest targets.For example,in the C-band PolSAR dataset,for the dihedral corner reflectors with 0°azimuth orientation angle,the average value of the estimated polarization orientation angle(?_e0)differs from the accrual value by only 2.7°.While for the dihedral corner reflectors with45°azimuth orientation angles,the average value of the?_e0is only 5.4°away from the actual value.In the decomposition results,the percentage of the double scattering is larger than 55.1%,the double scattering was always stronger than other scatterings.Therefore,although the dihedral corner reflectors have different azimuth orientation angles,they can be correctly delineated according to the dominant double scattering.In addition,the algorithm had only four components,the execution complexity of the algorithm was low,and each component had a clear physical meaning,the interpretation of the decomposition results was straightforward.(3)To analyze the scattering characteristics of and classify the azimuthal asymmetric forest targets in P-band PolSAR datasets,the Bragg scattering model was used to analyze the relationship between the scattering characteristics of the ground and the local incidence angle of radar.The azimuthal symmetry of the ground deteriorates with the increase of the radar local incidence angle.Then an eigenvalue/eigenvector-based PolSAR decomposition algorithm without the azimuthal symmetry assumption was proposed to extract the topography-related scattering of forest targets in P-band PolSAR datasets,and then the azimuthal asymmetry of the topography was eliminated by using the depolarization method,and the azimuthal symmetry of the forest targets was reconstructed.The effectiveness of the algorithm was verified by multiple AIRSAR and UAVSAR P-band PolSAR data.The algorithm can increase the volume scattering of forest targets in the decomposition results of the P-band PolSAR datasets by more than25.0%,and the recognition rate by more than 28.3%.