An Adaptive PolSAR Tomography Method Based on Scattering Mechanism Classification

Synthetic aperture radar (SAR) tomography is an essential method for acquiring data used in the reconstruction of urban 3-D models. With advancements in SAR technology, polarization has been integrated into tomography for accurately identifying and locating target structures. Recent studies focus on independently enhancing the ability of compressed sensing (CS) or spectral estimation methods to utilize polarimetric data. Since urban areas contain both small-scale man-made objects and large-scale homogeneous or distributed targets, using either of these two methods alone cannot balance accuracy with the preservation of 3-D details. To address this, this study presents an adaptive polarimetric SAR (PolSAR) tomography method based on scattering mechanism classification, which leverages the strengths of both CS and spectral estimation techniques. In this study, we focus on the inversion of a single scatterer, without any attempt to separate multiple scatterers. The proposed method classifies pixels into four categories based on their scattering mechanisms, typically corresponding to man-made and distributed targets. The processing algorithms for CS and spectral estimation are then selected automatically based on the classification. For the pixels processed using spectral estimation, an adaptive window is also employed to compute the covariance matrix. Furthermore, the method employs an optimal polarization basis projection technique to effectively utilize polarization information during the tomography process. Experimental results show that this approach significantly improves both reconstruction accuracy and structural preservation.