摘要
针对极化合成孔径雷达(SAR)地物分类难以大批量标定准确率的问题,提出了一种利用光学遥感图像标定极化SAR分类准确率的方法。首先,采用快速区域卷积神经网络(Fast-RCNN)方法对光学遥感图像进行地物分类;然后,应用基于感兴趣区域的分布式目标异源图像匹配方法,建立光学遥感图像与SAR图像的对应关系;最后,以光学遥感图像的分类结果估计极化SAR分类的准确性。实验结果表明:光学遥感图像的分类结果可以较好地估计极化SAR分类准确性,该方法具有重要的工程应用价值和现实意义。
In order to solve the problem that the polarization synthetic aperture radar(SAR) classification is difficult to estimate the accuracy, a method for estimating the accuracy of polarization SAR classification by optical remote sensing image is proposed. The optical remote sensing image is classified by the fast region-based convolutional neural network(Fast-RCNN) method, and then the region of interest-based distributed target heterogeneous image matching method is applied to establish the correspondence between the optical remote sensing image and the SAR image. The experimental results indicate that the classification results of optical remote sensing images can better estimate the accuracy of polarization SAR classification, showing important engineering application value and realistic meaning.
作者
刘川
尹奎英
LIU Chuan;YIN Kuiying(Nanjing Research Institute of Electronics Technology,Nanjing 210039,China)
出处
《现代雷达》
CSCD
北大核心
2020年第7期99-102,共4页
Modern Radar
关键词
极化合成孔径雷达
地物分类
深度学习
异源图像匹配
polarimetric synthetic aperture radar
terrain classification
deep learning
heterogeneous image matching
