CMA-ES for Autofocus in ISAR Images

Machine learning models are increasingly applied to radar imaging. Motion compensation and autofocusing in inverse synthetic aperture radar (ISAR) imaging is a crucial part and of particular relevance to achieve high-resolution images. Several algorithms have been employed to estimate the correcting phase shift, either directly or indirectly. This paper explores the application of covariance matrix adaptation evolution strategy (CMAES) for phase reconstruction in ISAR imaging and estimating the phase directly in real measured data. As a fitness function, the k space entropy is used. On this basis, the results are compared to those of particle swarm optimization (PSO) and multiple scatterer algorithm (MSA). CMA-ES is also applied after using MSA to determine whether additional improvements can be achieved.