Coherent radar imaging based on compressed sensing

Abstract

High-resolution radar images in the horizontal spatial domain generally require a large number of different baselines that usually come with considerable cost. In this paper, aspects of compressed sensing (CS) are introduced to coherent radar imaging. We propose a single CS-based formalism that enables the full three dimensional (3D)—range, Doppler frequency and horizontal spatial (represented by the direction cosines) domain—imaging. This new method can not only reduce the system costs and decrease the needed number of baselines by enabling spatial sparse sampling, but also achieves high-resolution in the range, Doppler frequency, and horizontal space dimensions. Using an assumption of point targets, a 3D radar signal model for imaging has been derived. By comparing numerical simulations with the fast Fourier transform (FFT) and maximum entropy (ME) methods at different signal-to-noise ratios (SNRs), we demonstrate that the CS method can provide better performance in resolution and detectability given comparatively few available measurements relative to the number required by Nyquist-Shannon sampling criterion. These techniques are being applied to radar meteor observations.