Fast Sparse Bayesian Learning Method for Scanning Radar Super-Resolution Imaging

Abstract

The high azimuth resolution and low computational complexity are significant for scanning radar super-resolution imaging. In this paper, we propose a fast sparse Bayesian learning method with truncated singular value (FSBL-T) to quickly improve the azimuth resolution. Firstly, we truncate the small singular value of the convolution matrix to rewrite the echo signal. Secondly, we introduce the Gaussian-Gamma distribution to model the echo noise and sparse target. Then, the expectation maximization method is utilized to estimate the imaging parameters. To reduce the computational load, we introduce Kailath Variant conversion to reduce the size of the inverse matrix. Finally, the simulations illustrate that the FSBL-T method can effectively enhance the azimuth resolution without reducing the imaging accuracy.

Keywords

• Scanning radar
• super-resolution imaging
• sparse Bayesian learning
• low computational complexity
• truncated singular value