Synthetic Aperture Radar (SAR) automatic target recognition is an important technology in remote sensing, and deep learning has greatly improved its performance. However, SAR images, which belong to the microwave vision spectrum and reflect the target’s backscattering, differ fundamentally from optical images. This difference creates challenges when directly applying optical-based deep learning techniques to SAR data, such as poor interpretability and a tendency to overfit. To overcome these issues, integrating SAR-specific characteristics into deep learning methods has become a key research focus. However, there is a lack of clear guidance on how to effectively consider these characteristics, limiting the effectiveness of recognition systems. This review answers three important questions: (1) What are the unique characteristics of SAR images? (2) How can these characteristics be effectively modeled? (3) How can SAR target features be integrated into deep learning models? By addressing these questions, this review aims to support the development of more reliable and interpretable SAR target recognition systems.