A Matrix Pencil Algorithm Based Multiband Iterative Fusion Imaging Method

Abstract

Multiband signal fusion technique is a practicable and efficient way to improve the range resolution of ISAR image. The classical fusion method estimates the poles of each subband signal by the root-MUSIC method, and some good results were get in several experiments. However, this method is fragile in noise for the proper poles could not easy to get in low signal to noise ratio (SNR). In order to eliminate the influence of noise, this paper propose a matrix pencil algorithm based method to estimate the multiband signal poles. And to deal with mutual incoherent between subband signals, the incoherent parameters (ICP) are predicted through the relation of corresponding poles of each subband. Then, an iterative algorithm which aimed to minimize the 2-norm of signal difference is introduced to reduce signal fusion error. Applications to simulate dada verify that the proposed method get better fusion results at low SNR.

Improved Microwave Imaging by Wavenumber Domain Multiband Data Fusion

“…Then the full band signalsS(m), m = 0, 1, • • • , M − 1, can be estimated, where M is the number of the samples in the full band with sampling intervals of ∆k. To refine the estimation of the full band signals, the similar iterative scheme as in [4] is applied.…”

“…The multiband data are modeled with autoregressive (AR) models or autoregressive moving average (ARMA) models over a wide bandwidth according to the scattering behaviors of canonical scatterers. Then the signal models can be estimated with root MUltiple SIgnal Classification (MUSIC) algorithm [1], [3], matrix pencil aproach [4] singular-value decomposition (SVD) [2], sparse Bayesian learning algorithm [5] and support vector machine [6]. In addition, a fusion method that combines all-phase fast Fourier transform (apFFT) and iterative adaptive approach [8] was proposed to fuse the dechirped multiband signals, which is more dedicated to the linear frequency modulated (LFM) signals.…”

“…By modeling the wavenumberdomain signals in each radial direction at different subbands as a superposition of damped/undamped exponential functions, the multiband fusion problem is converted to parameter estimation of the exponential components. Similar to [4], the signal model in each radial direction is estimated by using the matrix-pencil approach for signal extrapolation in the frequency gap between different frequency subbands. Finally, coherently processing the available data and extrapolated one, a resolution-enhanced image can be obtained.…”

Sub‐band mutual‐coherence compensation in multiband fusion ISAR imaging

“…Researchers tend to adjust the signal processing approach to achieve a similar performance, especially when upgrading existing radar systems [3]. Multiband fusion imaging technology can effectively improve the range resolution of the radar by fusing the target frequency responses in different sub‐bands to achieve a higher bandwidth in terms of the signal level without increasing the hardware cost [1, 4–8].…”

“…The compensation accuracy is limited by the sub‐band bandwidth, and the search operation is a heavy burden in real‐time processing. In [6, 13, 14], the sub‐band poles, which correspond to the target scatterers of each sub‐band, were utilised for ICP coherence compensation. The modified root multiple signal classification (root‐MUSIC) method [1] was adopted to obtain pole estimations, then the phase differences of the poles and amplitudes between each sub‐band were utilised to estimate the ICPs.…”