A Study of Recovering Quad-Polarimetric Information from SLC/MLC Compact-Polarimetric SAR Data via CNNS

Quad polarimetric (quad-pol) Synthetic Aperture Radar (SAR) has unique advantages to record the detailed backscattering information, which significantly enhances the interpretation quality of objects. However, the quad-pol SAR system’s intricate nature imposes constraints on its swath coverage and spatial resolution. A promising solution to this challenge lies in the recovery of quad-pol information from Compact-Polarimetric (CP) SAR data. In this paper, recovery methods based on the convolutional neural network (CNN) are introduced, for both single-look complex (SLC) and multi-look complex (MLC) SAR data. It also presents a comparative performance analysis of these recovery techniques across two data types, emphasizing the feasibility and efficiency of two approaches in overcoming the limitations of quad-pol SAR systems while maintaining high data quality.