Automatic target recognition of SAR images based on Transformer

In order to solve the overfitting problem of synthetic aperture radar automatic target recognition (SAR-ATR) at small samples scenarios, this paper proposes a non-subsampling laplace pyramid decomposition (NSLP) based vision transformer (VIT) model, namely NSLP-VIT, for SAR-ATR task. First, the SAR images are preprocessed by feature decomposition via NSLP; Second, a three-channel image is merged as the input of the network using three feature sub-maps; Finally, the preprocessed SAR images are as inputs for VIT network training. The experimental results on Moving and Stationary Target Acquisition and Recognition (MSTAR) data sets show that the proposed method can effectively solve the overfitting problem of SAR-ATR when the data is scarce.