3D ISAR Image Reconstruction of Ground Vehicles Using 3D SL0 on Under-Sampled Aperture Data

This paper considers the three-dimensional (3D) reconstruction of inverse synthetic aperture radar (ISAR) imagery from under-sampled apertures using compressive sensing. Previously, we implemented a 2D version of the smoothed
$\ell^{0}$
norm algorithm (2D SLO) on 3D data by creating a 2D data phase history matrix of frequency by aspect angle pairs. The advantage in this previous formulation is the ability to sense specific azimuth/elevation pairs. While the previous formulation includes a reduction in computer memory due to its use the the 2D SLO algorithm, here we have implemented a 3D SLO algorithm that greatly increases the computational efficiency by successive application of the ID SLO algorithm on individual dimensions. Results are provided using synthetic data for a scaled Chevrolet CK1500 pickup truck. We see that the 3D SLO algorithm maintained the integrity of the sparse signal reconstruction by successfully reconstructing nearly all of the dominant scattering centers, while being much faster than our previous approach due to significantly fewer matrix operations.