Azimuth super-resolution of forward-looking imaging based on bayesian learning in complex scene

Abstract Azimuth super-resolution is an efficient method to enhance the angular resolution of scanning radar in forward-looking area. The existing super-resolution methods are limited in a simple scene, which mainly consider a single clutter environment. For some complex scenes, the distribution of clutter is more complex than single clutter, such as sea-surface scene or the junction scene of sea-surface and ground. In this paper, we present a sparse Bayesian learning method to promote the azimuth resolution in a complex forward-looking scene. First, we use the Gaussian mixture model (GMM) to express the statistical character of the clutter and noise in the complex scene. Second, we present Laplace hierarchical prior as the prior information to model the sparse target. Then, the joint distribution of the clutter and target is derived as an optimized problem under the Bayesian framework. Finally, the solution is solved by the expectation maximization (EM) based maximum a posterior (MAP) method. The simulation and semi-real data results show that the proposed algorithm provides better angular resolution than traditional methods in complex scene.