Orthogonal Matching Pursuit-Sparse Bayesian Learning Based SAR Tomographic Inversion of Urban Area

Tomographic synthetic aperture radar (TomoSAR) enables high-precision three-dimensional (3-D) reconstruction of urban buildings. However, conventional algorithms require large data stacks and uniform baseline distribution. Compressed sensing (CS) based methods not only exhibit the super-resolving imaging ability, but also reduce the number of samples requirement for elevation recovery. In this letter, orthogonal matching pursuit-sparse Bayesian learning (OMP-SBL) method is proposed to the tomographic inversion of urban area from the complexvalued image dataset of Fucheng-1 synthetic aperture radar (SAR) satellite. In the proposed method, orthogonal matching pursuit (OMP) is firstly employed to compute the prior and obtain initial sparse solutions. Then this prior is further refined by sparse Bayesian learning (SBL) algorithm. Compared with spectrum estimation and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$L_{0}$</tex>-norm minimization based algorithms, the proposed method achieves enhanced super-resolution performance without setting hyper-parameters.