Generating dense and super-resolution ISAR image by combining bandwidth extrapolation and compressive sensing

Abstract

Recent developing compressive sensing (CS) theory indicates that it is possible to obtain precise recovery of a sparse signal from very limited measurements, which provides a new way for data acquisition and signal processing as nature signals usually involve some degree of sparsity. In this paper, we present an algorithm for inversed synthetic aperture radar (ISAR) imaging with super resolution by combining CS and bandwidth extrapolation (BWE) technique. For ISAR imaging, the backscattering field of target is usually contributed by a few strong scattering centers, whose number is much less than that of image pixels. Thus, CS is intuitively suitable for constructing super resolution ISAR image. According to CS theory, the number of extracted dominating scatterers relies on the signal length, which indicates that if only limited data is available, it is difficult to generate dense ISAR image robustly by CS, and some signal components tend to lose. To soften this constraint, BWE is combined with CS imaging to increase the degree of freedom of signal while preserving its coherence. A refined CS-based formation for ISAR image-resolution enhancement is then developed. Both real and simulated data experiments are performed to evaluate the proposed approach, and an example of using this technique demonstrates the enhanced image resolution in application of maneuvering target imaging.

Isar Imaging of Non-Uniform Rotation Targets With Limited Pulses via Compressed Sensing

“…The state-of-art multifunction radar system usually provides gapped and short interval observation, the corresponding sampling pattern in Doppler domain can be considered in the construction of the azimuth sparse dictionary [13], thereby the super-resolution image can be obtained via CS algorithm. Suppose the compressive sampling matrix is Φ, in the short aperture case, the action of Φ is equivalent to selecting partial rows of Ψ in series, thus Φ is a matrix consisting of partial sequential rows of an identity matrix.…”

“…Recently, the CS-based short aperture and sparse aperture ISAR imaging methods have already attracted extensive attention and made preliminary progress [10][11][12][13][14][15][16][17]. However, all the above-mentioned work are under the assumption that the target's rotation during the CPI is uniform, so the echoed signal in a range cell is composed of single frequency components, and the corresponding sparse dictionary is a discrete Fourier matrix.…”

A novel spaceborne SAR wide-swath imaging approach based on Poisson disk-like nonuniform sampling and compressive sensing

“…Although the computation cost of direct implementation increases from O(2M N ) to O(N 2 ) along with the decrease in the number of multiplications, as mentioned above, A is composed of uniform and nonuniform DFT bases, which means we can accelerate the matrix-vector multiplication efficiently by making use of FFT algorithm. By (11) we rewrite A H A s k−1 as…”

“…The new concept of compressive sensing (CS) states that an unknown sparse (or sparse under some transform-domain) signal can be recovered even from what appear to be highly sub-Nyquist-rate samples by solving an l 1 -optimization problem [4,5], thus offering the possibility of realizing wider swath for conventional single channel spaceborne SAR system for sparse scenes such as sea. CS theory can effectively reduce the system complexity in acquiring image technique and has been already applied to the radar imaging techniques [6][7][8][9][10][11]. In existing literature, the azimuth under sampling pattern in SAR imaging based on CS is typically designed as the random selection strategy, but this method cannot be applied to wide-swath imaging in spaceborne SAR since it is unable to ensure that the time interval between any two adjacent samples is smaller than the one under the Nyquist limitation.…”

Resolution enhancement for Doppler beam sharpening imaging

“…To overcome the limitation on sharpening ratio of conventional method based on FFT, we present a combined super-resolution algorithm with aperture extrapolation (AE-CSR), which includes two key procedures. First, an autoregressive (AR) model is established in the 'angular' domain, and the aperture extrapolation (AE) [13][14][15][16] technique is used to extrapolate some parts of the spectrum outside the bandpass of system by using the measured limited pulses. Second, we utilise the spectral estimation method rather than FFT to perform the Doppler analysis.…”