摘要
特征提取是合成孔径雷达(synthetic aperture radar,SAR)图像自动识别与分类中的重要环节.由于SAR图像有相干斑噪声及几何畸变等特性,一般网络模型难以提取到有判别性的特征.为增强特征提取能力,提高分类准确率,提出将注意力机制与胶囊网络结合的一种注意力胶囊网络模型.注意力机制可聚焦寻找具有重要局部信息的特征,在图像识别过程中抑制干扰特征,定位重要特征.胶囊网络可捕捉图像中目标的位置与空间关系,使提取到的SAR图像特征含有更多便于分类的重要信息.结果表明:文中方法对SAR图像分类数据集中运动和静止目标的获取与识别(moving and stationary target acquisition and recognition,MSTAR)有显著效果.
Feature extraction is a very important part of automatic recognition and classification of synthetic aperture radar(SAR)images.However,due to characteristics of coherent speckle noise and geometric distortions in SAR images,it is difficult for general network models to extract discriminative features.In order to enhance the feature extraction ability and improve the classification accuracy,this paper proposes an attention capsule network model,which combines the attention mechanism with the capsule network.The attention mechanism can focus on finding features with important local information,suppressing unwanted features in the image recognition process and locating important features.The capsule network can capture the position and spatial relationship of the target in the image so that the extracted SAR image features contain more important information that is conducive to classification.Experimental results show that the proposed method has achieved significant results on the MSTAR dataset.
作者
王权
温显斌
WANG Quan;EN Xianbin(Key Laboratory of Computer Vision and System,Ministry of Education,Tianjin University of Technology,Tianjin 300384,China;Tianjin Key Laboratory of Intelligence Computing and Novel Software Technology,Tianjin University of Technology,Tianjin 300384,China;School of Computer Science and Engineering,Tianjin University of Technology,Tianjin 300384,China)
出处
《天津理工大学学报》
2023年第6期42-48,共7页
Journal of Tianjin University of Technology
基金
国家自然科学基金(61472278)
天津市新一代人工智能重大专项(18ZXZNGX00150)
天津市自然科学基金(18JCYBJC84800)
天津市教委自然科学项目(2017ZD13,2017KJ255)。
关键词
自动目标识别
合成孔径雷达
注意力
胶囊网络
automatic target recognition
synthetic aperture radar
attention
capsule network
