基于跨域小样本学习的SAR图像目标识别方法  

SAR Image Target Recognition Based on Cross Domain Few Shot Learning

作　　者：史松昊 王晓丹[1] 杨春晓[1] 王艺菲[1] SHI Songhao;WANG Xiaodan;YANG Chunxiao;WANG Yifei(College of Air and Missile Defense,Air Force Engineering University,Xi’an 710051,China)

机构地区：[1]空军工程大学防空反导学院,西安710051

出　　处：《计算机科学》2024年第S01期453-459,共7页Computer Science

基　　金：国家自然科学基金(61876189,61703426,61806219)。

摘　　要：由于SAR图像获取难度大,可供研究的样本数量较少,解决有限样本条件下SAR图像目标识别问题成为业界公认的挑战。随着深度学习在计算机视觉领域的发展,衍生出了多种小样本图像分类方法,因此考虑采用跨域小样本学习范式解决小样本SAR图像目标识别问题。具体地,先在多个源域中训练得到不同域的特征提取器,而后通过知识蒸馏的方法获取一个通用的特征提取器,这里采用中心核对齐的方法,将提取的特征映射到一个更高维的空间,从而更好地区分原特征之间的非线性相似性;通过上一阶段获得的通用特征提取器提取目标域图像特征,最后采用原型网络的方法预测样本的类别。实验证明,该方法在缩减模型参数的同时,获得了88.61%的准确率,为解决小样本SAR图像目标识别问题提供了新的思路。Due to the difficulty in acquiring SAR images and the scarce number of samples available for research,solving the SAR image target recognition problem under few shot conditions has become a community-recognized challenge.With the development of deep learning in the field of computer vision,a variety of few-shot image classification methods have been derived,so a cross-domain few-shot learning paradigm is considered to solve the few-shot SAR image target recognition problem.Concretely,the feature extractors of different domains are first trained in multiple source domains,while a generalized feature extractor is obtained by knowledge distillation.In this stage,the central kernet alignment method is used to map the extracted features to a higher dimensional space,so as to better distinguish the nonlinear similarity between the original features.Then the target domain image features are extracted by the generalized feature extractor obtained in the previous stage.Finally,a prototype network approach to predict the class of the sample.The experiment proves that the method obtains 88.61%accuracy while reducing the model parameters,which provides a new method for solving the target recognition problem of SAR images with scarce samples.

关 键 词：深度学习 元学习 跨域小样本学习 SAR图像目标识别 知识蒸馏 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TN957.52[自动化与计算机技术—控制科学与工程]