Overcoming polar-format issues in synthetic aperture radar multichannel autofocus

摘要

Multichannel autofocus (MCA) is a subspace-based autofocus method for solving the defocusing problem in synthetic aperture radar. In addition to the one-dimensional (1D) defocusing assumption, MCA assumes that the perfectly focused image has a low-return region, which is naturally guaranteed by the spatially limited nature of the radar antenna footprint. In theory, MCA yields better or even perfect solutions compared to other autofocus methods. However, the authors have discovered that MCA is far more sensitive to violation of the 1D defocusing assumption compared to other methods; in fact, MCA is unsuitable for even fairly small data-collection angles. Fortunately, this problem can be solved if they reverse the order of two steps in the image formation process and apply MCA in a domain where the defocusing effect is one dimensional. The distorted version of the image, obtained by inverse Fourier transforming the polar-grid data without further interpolation, contains regions satisfying a low-return assumption, but the region of low return must be carefully specified for best performance. They present simulation results of the proposed method, reversed-step MCA, for various ranges of look angles and discuss the selection of low-return constraints.