Norm Optimization using Machine Learning Approach for Autofocus in mmWave SAR Imaging

Abstract

Due to the shorter wavelength of mm-wave SAR for higher resolution, high-frequency phase error (HPE) can be generated even by small vibration of antenna phase center and distortion to SAR image becomes significant. For the problem, the machine learning approach can be utilized in SAR autofocus by classifying images and optimizing the objective function for autofocus. A hybrid form of L1/ L2-norm is adapted to the range compressed data corresponding to the input of the autofocus taking advantage of the convergence speed and the stability. Its convergence feature is analyzed and demonstrated in the simulation.