Compressive sensing for inverse synthetic aperture radar imaging

In the last few decades, the increasing demand for high-resolution images in applications such as automatic target recognition (ATR) and automatic target classification (ATC) for surveillance and homeland security has attracted the attention of the research community worldwide [35]. In this scenario, inverse synthetic aperture radar (ISAR) imaging techniques have been widely studied since they allow to obtain high-resolution images of moving targets by means of Fourier-based imaging methods [11,53]. In particular, the transmission of large bandwidth signals and the coherent integration of the received echoes from different aspect angles are the key aspects to obtain images with fine resolution both in range and cross-range. By the way, in some cases, the transmitted bandwidth and/or the coherent processing time (CPI) might be not enough for the desired resolution. In addition, system malfunctioning, transmission disruption, or data compression may be the causes of missing samples, both in the frequency and/or in the slow-time domain. In such cases, conventional Fourier-based imaging methods, such as the range-Doppler (RD) method, suffer from coarse resolution and distortions [11]. To enhance the results obtained via RD methods, different super-resolution (SR) techniques have been studied in the literature, such as apodization 106techniques [14,15] and bandwidth extrapolation (BWE) techniques [56]. The main disadvantages of such SR methods are that their performance is very sensitive to noise and clutter and that these methods generally computationally expensive.