Urban Damage Mapping With Reconstructed Quad-Pol SAR Data From Dual-Pol SAR mode

Polarimetric synthetic aperture radar (PolSAR) has the advantage of all-day and all-weather work, which can obtain high-resolution radar images and polarimetric scattering matrix of the target. The dual polarimetric SAR can obtain a wider observation swath with lower energy budget than quad-pol SAR. In this paper, the pseudo quad-pol SAR data is reconstructed from dual polarization SAR data based on the proposed deep convolutional neural network. The urban damage level mapping index is established based on the co-polarization coherence pattern and the reconstructed quad-pol SAR data. The experiments have been conducted on the multi-temporal ALOS/PALSAR data from the March 11 Eastern Japan earthquake. Compared with the existing quad-pol SAR data reconstructed method, the proposed method has superior reconstruction performance. The quantitative experiment results indicate the proposed method can effectively achieve urban damage level mapping based on the dual polarization SAR data.