High Resolution Large Elevation Synthetic Aperture Tomographic SAR Imaging With Classifier-Based Imaging Architecture

Conventional Tomographic SAR (TomoSAR) elevation processing assumes that the wavefronts from different radar apertures are approximately parallel within the observation range, meaning that all targets in space with the same slant range overlap on the same iso-range ring. We refer to the above assumption as the parallel wavefront assumption. However, the TomoSAR elevation resolution is positively correlated with the synthetic aperture angle in the elevation dimension. As the elevation synthetic aperture angle increases, the parallel wavefront assumption becomes invalid, resulting in significant deviations of the TomoSAR imaging model. In this paper, we propose a new TomoSAR imaging processing architecture named “Elevation Aperture Signal-extraction, Target-classifier, and Estimation Reconstruction” (EASTER) for large elevation synthetic aperture TomoSAR imaging. The core mechanism of EASTER lies in utilizing a classifier to address the target localization problem of TomoSAR imaging, thereby avoiding sparse signal processing based on the parallel wavefront assumption. Specifically, “Elevation Aperture Signal-Extraction” involves extracting the multi-baseline TomoSAR echo from different single look complex (SLC) images using range-doppler (RD) positioning to identify signals that share the same range-azimuth trajectory as a specified grid point. The “Target-Classifier” refers to the use of a target classifier to classify the elevation signals and determine whether a target exists at the specified grid point. The “Estimation Reconstruction” entails recovering the amplitude and phase information of the targets based on the coherence principle. Simulation and real UAV data experiments conducted performance analysis of the proposed method compared to BP and CS methods, based on 3D Structural similarity index measure (SSIM) and chamfer distance (CD) metrics. The results confirmed that the proposed method can more accurately retrieve the true structural features of the target.