An ISAR Autofocus Imaging Algorithm Based on FCN and Transfer Learning

Inverse Synthetic Aperture Radar (ISAR) imaging targets are mostly non-cooperative targets and motion compensation is the key to require high quality of the ISAR imaging. Fully Convolutional Networks (FCN) have a powerful ability to extract and represent data abstract features. In this paper, an ISAR autofocusing algorithm based on FCN is proposed, which is combined with transfer learning. Firstly, a large number of simulation data sets with different phase errors were constructed and trained to make FCN have the phase compensation capability. Secondly, the translational phase compensation ability of FCN is improved by utilizing transfer learning to combine the common characteristics between the simulated and real data. Finally, analyze the performance of traditional methods and FCN objectively. The real data processing results show that the performance of FCN phase compensation algorithm is better than traditional algorithm, which verifies the effectiveness of the proposed method and the superiority compared with the traditional method.