Compressive radar with off-grid targets: a perturbation approach

Abstract

Compressed sensing (CS) schemes are proposed for monostatic as well as synthetic aperture radar (SAR) imaging with chirped signals and ultra-narrowband (UNB) continuous waveforms. In particular, a simple, perturbation method is developed to reduce the gridding error for off-grid targets. A coherence bound is obtained for the resulting measurement matrix. A greedy pursuit algorithm, support-constrained orthogonal matching pursuit, is proposed to take advantage of the support constraint in the perturbation formulation and proved to have the capacity of determining the off-grid targets to the grid accuracy under favorable conditions. Alternatively, the locally optimized thresholding is proposed to enhance the performance of the CS method, basis pursuit. For the advantages of higher signal-to-noise ratio and signal-to-interference ratio, it is proposed that Spotlight SAR imaging will be implemented with CS techniques and multi-frequency UNB waveforms. Numerical simulations show promising results of the proposed approach and algorithms.