Research Of Urban Building Target Recognition Based On Compact Polarimetric SAR Imagery

The concept of Compact Polarimetric Synthetic Aperture Radar(CP SAR)in recent years has received more and more attention.CP SAR is essentially a dual-pol system.The design complexity of the system is lower than that of a quad-pol system,but the imaging width of the ground is larger than that of a quad-pol system.The CP SAR has shown that it is almost equivalent to a quad-pol system in many applications.Due to the significance of urban targets and the complexity of scattering characteristics,urban areas have always been the focus of microwave remote sensing applications.Existing CP SAR decomposition algorithms,the decomposition results are prone to overestimate volumetric scattering.Because the urban buildings not perpendicular to the direction of radar wave,the azimuths angle will further overestimate the volumetric scattering,which cause misclassification.In this paper,the following work is carried out on the problem of CP SAR system for urban target recognition.1.In this paper,the Souyris reconstruction algorithm and the Nord reconstruction algorithm are derived and implemented in detail.Simulated three CP SAR mode data through quad-pol SAR data.The reconstruction quality of the key parameter of cross-pol component is presented.2.The research implement two typical methods of m-? decomposition,m-? decomposition,which are directly finish decomposition by using the polarization characteristic parameters.Their decomposition results in the target area containing urban buildings has been analyzed.Using these two decomposition methods and compare with Pauli decomposition results of the quad-pol data,we found that the problem of overestimation of volumetric scattering in the CP SAR decomposition is due to only regard the depolarized component as volumetric scattering mechanism.3.With the help of the characteristics of ?/4 CP mode,an improved model-based decomposition method is proposed.Due to the complexity of the scattering characteristics of urban area,urban area should not be regarded as only composed of fully polarized components,but composed of multiple scattering components.Using polarization characteristic parameters and the Freeman-Durden volumetric scattering model,increase the proportion of double-bounce scattering components in urban areas.Due to the assumption of reflection symmetry,the correlation coefficients of the co-polarized and cross-polarized scattered components are set to zero.However,this item has the function of compensating for the azimuth angle effect of urban buildings in the study of quad-pol SAR data,so it is more difficult to identify the large azimuth urban area using CP SAR data.In this paper,an improved model is proposed using the relationship between Stokes parameters and the quad-pol coherency matrix.We used data from different platforms and operating bands,including ALOS PALSAR data and GAOFEN-3 data in the San Francisco area of the United States,and E-SAR data in the Oberpfaffenhofen of Germany.ALOS PALSAR experiment show that our method can effectively improve the ratio of double-bounce scattering components in urban areas with large azimuth angles,and obtain better decomposition results than existing metheds.In the small azimuth urban area,compared with the model-based method and the pseudo fully polarized reconstruction decomposition method,the double-bounce scattering energy is increased by 16.4% and 40.7%,respectively.In the urban area with a large azimuth angle,compared with the model-based method and the decomposition method after pseudo fully polarization reconstruction,the double-bounce scattering energy is improved by 7.3% and 21.6%.In the view of classification,the false dismissal probability index dropped more,indicating the ability to detect targets for urban buildings.