SAR Image Classification Algorithm Combining Data Expansion and Transfer Learning

SAR images have the advantage of being unaffected by weather and terrain. It has become an important research direction in microwave imaging field. However, most traditional image classification methods are based on shallow features and rely on manual extraction, which is difficult to meet the current technical requirements. With the gradual maturity of SAR imaging technology, how to effectively use SAR image to complete the task of recognition and classification has become an urgent problem to be solved. In this paper, a classification method combining data expansion and transfer learning is proposed to expand the dataset in the original domain, enrich the diversity of samples, introduce the idea of transfer learning, break the limitations of dataset, and improve the training efficiency of the model. Finally, a large number of experiments are carried out on MSTAR image dataset to verify the feasibility of this scheme. The experimental results show that the proposed method can accelerate the network convergence and improve the classification performance and training efficiency of the model.