Target Representation and Classification with Limited Data in Synthetic Aperture Radar Images

Recently, with the introduction and development of deep learning based detection and classification methods, various applications in optical images have been put into practice. However, few automatic target recognition (ATR) approaches in Synthetic Aperture Radar (SAR) images can be used practically. Two reasons underlying it are the complicated imaging mechanism of SAR and limited sample data for optimizing the models. This paper focus on target representation and classification with limited data based on zero-shot learning (ZSL) and fewshot learning (FSL), and provides a comprehensive investigation of existing ZSL/FSL algorithms.