<title>3D SAR imaging using a hybrid decomposition superresolution technique</title>

Abstract

A technique to form super-resolved 3D Synthetic Aperture Radar (SAR) images from a limited number of elevation passes is presented in this paper. This technique models the environment as containing a finite number of isotropically radiating, frequency independent point scatterers in Additive White Gaussian Noise (AWGN), and applies a hybrid super-resolution method that yields the Maximum Likelihood (ML) estimates of scatterer strengths and resolves their locations in the data deficient dimension well beyond the Fourier resolution limit.

Three-dimensional target modeling with synthetic aperture radar

“…A second technique for high-resolution spectral estimation of radar scatterer amplitudes and altitudes is presented by Walter S. Kuklinski and Andrea L. Kraay [8] . This technique aims to provide the Maximum Likelihood (ML) estimate of scatterer strengths and locations based on a decoupled least-squares process of amplitude estimation.…”

Three-Dimensional Target Modeling with Synthetic Aperture Radar

“…The implementation of the decoupled least-squares technique in simulation was based on the mathematical description of the technique by Kuklinski and Kraay [8] and the flow chart and equations given in Section 3.4.2. The technique was altered slightly for the sake of testing it against some of the parameters, but the Maximum Likelihood method for the estimation of reflector locations and contributions is the same.…”

“…By least-squares, the maximum likelihood estimate of scatterer amplitudes, equivalent to an estimated version of matrix a, is [8] ˅ {ˀˀ { #ˀˑ (3.13) Overall, the primary inputs required to the system are SAR frequency data samples in the z domain for each (x,y) location. The noise is already accounted for in the Maximum…”

“…The estimates altogether form an estimated version of the frequency domain data signal referred to as ˟ ӂ { { such that the estimates adhere to a least-squares model and the cost function C is minimized: [8] ˕ ˗ Ӛ ˟{ { . ˟ ӂ { { $ ӛ (3.11)…”

“…A new matrix V is defined, which encompasses the v matrix for all samples of ω z . The noise model used is a zero-mean complex Gaussian white noise vector, and the overall system model, in expanded matrix form, is [8] ˑ…”