Review of Hybrid Model Used in SAR Target Recognition

SAR target recognition has a solid theoretical foundation and broad application prospects in both civil and military fields. Model-based target recognition generally includes feature extraction and classifiers. The recognition speed is faster and the recognition effect is better under limited sample conditions. However, it needs to rely on feature analysis and designe manual features. On the high-dimensional logic, feature selection and feature combination are also difficult. Recognition methods based on deep learning generally include convolutional neural networks, deep belief networks, encoders, etc. and have high recognition accuracy. However the methods are highly dependent on the amount and distribution of data. In the existing research, part of the research involves the combination of methods based on model and methods based deep learning. This article analyzes and reviews the existing hybrid model combining the two methods on SAR target recognition.