Application of <italic>H</italic>/&#x1FB1; Decomposition to Limited and Dual-Polarimetric 3D SAR Data of Civilian Vehicles

Abstract

This paper investigates the feasibility of scattering mechanism identification from limited synthetic aperture radar (SAR) data for civilian vehicles. Wide-angle 3D SAR imaging is considered with limitations in both frequency/look-angle and polarization samples. Accordingly, the two main problems, namely, sparse reconstruction of wide-angle data and identification of scattering mechanisms using two polarization channels, are jointly addressed. A methodology involving compressed-sensing (CS) imaging, processing of sub-aperture images, and application of $H/\bar {\alpha }$ decomposition to the dual-circular polarization (DCP) mode is proposed. The 2D and 3D maps of entropy $(H)$ and alpha-angle $(\bar {\alpha })$ parameters and $H/\bar {\alpha }$ classification results are evaluated by using simulation and the real GOTCHA dataset. The approach is tested with a complementary situation that consists of back-projection (BP) imaging of complete data plus decomposition of full-polarimetric (FP) data. A good correlation between full-available and most-limited cases, i.e., BP-FP vs. CS-DCP, is observed especially for the $\bar {\alpha }$ signatures. The results indicate a reasonably accurate retrieval of canonical mechanisms from a very small subset, i.e., about 0.23% of the total samples of each DCP channel.

Keywords

• Polarimetric SAR
• compact polarimetry
• compressed sensing
• 3D circular SAR imaging
• eigenvector decomposition