Enhanced Polarimetric Radar Imaging Using Cross-Channel Coupling Constraints

Data for a scene of interest may be collected over multiple polarization channels. In the case of polarimetric synthetic aperture radar images, regularization techniques are typically applied independently to each polarimetric channel. However, independent processing does not account for cross-channel coupling and may corrupt the polarimetric information in the signals. Recent consideration of joint enhancement techniques has shown promising results for multi- channel datasets with similar regions of signal magnitude and/or phase. However, in the case of polarimetric SAR data, scattering may be present in some channels and not in others. This thesis mathematically formulates multi-channel sparse imaging for polarimetric radar data using a joint enhancement algorithm to enforce sparsity and polarimetric coupling constraints. Two candidate functional relationships are derived to describe polarimetric coupling among received signal channels: one convex function and one non-convex function. These functions are reformed as optimization constraints. Then, an optimization problem is constructed to maintain signal fidelity, enforce sparsity, and preserve interchannel coupling. An iterative dual gradient descent algorithm is used to alternatively calculate updated scene estimates for each channel and the maximizing Lagrange multipliers for each coupling constraint. Results are found for several polarimetric SAR datasets. Jointly enhanced images are compared with corresponding images found through independent enhancement, taking into consideration signal fidelity, sparsity, polarimetric preservation, and scattering classification. Overall, the jointly enhanced image channels display significantly better polarimetric preservation compared to the corresponding independently restored image channels. More research is needed to understand how improved polarimetric preservation can be used to improve target classification