Improved Microwave Imaging by Wavenumber Domain Multiband Data Fusion

Abstract

A matrix-pencil based wavenumber-domain multiband fusion approach is proposed for improved microwave imaging. The suggested fusion approach is based on the fact from the Born approximation that the wavenumber-domain signal in a given scattering direction can be represented as a sum of the same number of contributions over the whole bandwidth. Utilizing this fact, the wavenumber-domain signal in each radial direction can be modeled as a superposition of damped/undamped exponential functions, where measured multiband data are part of the observations of the whole-band signal. Then the multiband signal fusion is addressed by estimating a unified signal model over the while bandwidth with the matrix pencil approach in the wavenumber domain. By using the estimated signal model, the missing data in the frequency gap can be extrapolated, thus synthesizing an equivalent wideband signal spectrum. After an inverse Fourier transform, the synthesize spectrum leads to a focused image with enhanced resolution. Compared to the existing frequency-domain method, the proposed fusion approach can be used for radar imaging with the signals acquired by either collocated or non-collocated arrays in different frequency bands. The effectiveness and accuracy of the proposed fusion approach are demonstrated through some numerical simulations.