Edinburgh Napier University

  Semisupervised learning in synthetic aperture radars (SARs) is one of the research hotspots in the field of radar image automatic target recognition. It can efficiently deal with challenging environments where there are insufficient labeled samples and large unlabeled samples in the SAR dataset. In recent years, consistency regularization methods in semisupervised learning have shown considerable improvement in recognition accuracy and efficiency. Current consistency regularization approaches suffer from two main shortcomings: first, extracting all of the relevant information in the image target is difficult owing to the inability of conventional convolutional neural networks to capture global relational information; second, the standard teacher–student regularization methodology causes confirmation biases due to the high coupling between teacher and student models. This article adopts an innovative trimodel semisupervised method based on attention-augmented convolutional networks to address the aforementioned obstacles. Specifically, we develop an attention mechanism incorporating a novel positional embedding method based on recurrent neural networks and integrate this with a standard convolutional network as a feature extractor, to improve the network's ability to extract global feature information from images. Furthermore, we address the confirmation bias problem by introducing a classmate model to the standard teacher–student structure and utilize the model to impose a weak consistency constraint designed on the student to weaken the strong coupling between the teacher and the student. Comparative experiments on the Moving and Stationary Target Acquisition and Recognition dataset show that our method outperforms state-of-the-art semisupervised methods in terms of recognition accuracy, demonstrating its potential as a new benchmark approach for the deep learning and SAR research community.