Attribute-Guided Multi-Scale Prototypical Network for Few-Shot SAR Target Classification
Siyuan Wang, Yinghua Wang, Hongwei Liu, Yuanshuang Sun
Abstract
Few-shot synthetic aperture radar (SAR) target classification has received more and more attention in recent years, where most of the existing methods have applied existing networks designed for natural images to SAR images, ignoring the special characteristics of SAR data. Therefore, in this paper, we propose an attribute guided multi-scale prototypical network (AG-MsPN) combined with subband decomposition for few-shot SAR target classification, aiming to learn more discriminative features from a few labeled data. Since the SAR images are essentially complex-valued images containing both amplitude and phase information, we employ the subband decomposition on complex-valued SAR images to explore the backscattering behavior variation of targets to get more complete descriptions of targets. Then, considering the complementary information extracted by different convolutional layers, based on prototypical network, a multi-scale prototypical network is proposed to aggregate the features of different layers to enhance the discrimination of feature representations, thus relieving the problem of intraclass diversity and inter-class similarity. Besides, we devise the prior binary attributes of SAR targets and present the attribute classification module to further strengthen the performance of multi-scale prototypical network under the joint supervision of the label and prior attribute information. We demonstrate the effectiveness of our proposed AG-MsPN on the Moving and Stationary Target Acquisition and Recognition (MSTAR) benchmark dataset and our method surpasses many other existing methods in the few-shot cases.