Super-resolution Imaging Method for Forward-looking Scanning Radar Based on Two-layer Bayesian Model

Abstract

Forward-Iooking scanning radar imaging is achieved by inversion of the matrix. However, the antenna measurement matrix pathology amplifies the noise of the inversion process, making direct inverse convolution imaging a pathological issue. Therefore, it is necessary to convert this problem into a benign one. A Bayesian model consisting of two layers is utilized to model the forward-looking super-resolution imaging noise and the target. In the first layer, the noise likelihood model is constructed as a normal distribution, while in the second layer, it is constructed as a gamma prior distribution conjugate to the normal distribution. The likelihood model parameters are updated through autonomous iterations, which can effectively fit the actual scenario noise. This approach allows for the recovery of the target scattering coefficients with good beam sharpening ability, even in situations with low signal-to-noise ratios (SNR). Simulation experiments have demonstrated the effectiveness of the proposed method.