摘要
过去二十年中,合成孔径雷达(Synthetic Aperture Radar,SAR)自动目标识别已经受到越来越多的关注。许多有监督特征学习算法被提出来,而且应用到合成孔径雷达的自动目标识别中。本文采用了无监督学习算法——K均值(K-means)聚类算法,通过分块自编码和优化接受域参数进行SAR图像特征学习,从而自动学习到无标签数据中鉴别性特征,并将所提取特征用于SAR图像目标识别中。然而,无监督学习一般对训练数据量有较高要求,因此,我们提出了两种数据增强方法,通过旋转目标物体的方位角,以及在原始图像上增加随机值,来获得更多可以用来训练模型的数据,使模型可以学习具有多样性的特征,达到提高识别效果的目的。采用公开的MSTAR数据库进行实验验证,结果表明所提方法可达到96.67%的主流识别率。
In the past two decades,Synthetic Aperture Radar( SAR) has received more and more attention. Many supervised algorithms have been proposed for automatic target recognition. In this paper,an unsupervised learning algorithm named K-means clustering is adopted,By coding the patch of the image and adjusting the receptive field size of the patch,the distinguish features can be learned from the input data for automatic target recognition. Furthermore,more training data is generated by using the data augmentation via rotating the azimuth of the target and adding a random integer to the original image for improving the training performance of the proposed algorithm. Experimental results on the public MSTAR database have shown that the proposed method can achieve the state-of-art accuracy,which is 96. 67%.
出处
《信号处理》
CSCD
北大核心
2017年第3期452-458,共7页
Journal of Signal Processing
基金
国家自然科学基金(61372193)
广东高等学校优秀青年培养计划项目(SYQ2014001)
广东省特色创新类项目(2015KTSCX143
2015KTSCX145
2015KTSCX148)
广东省青年创新项目(2015KQNCX172)
关键词
合成孔径雷达
无监督学习
K均值特征
数据增强
synthetic aperture radar
unsupervised algorithm
K-means feature
data augmentation