刘畅,
中国科学院电子学研究所 ??北京 ??100190
中国科学院大学 ??北京 ??100049
基金项目:国家自然科学基金(61471340)，国家重点研发计划(2017YFB0503001)
详细信息
作者简介:肖东凌(1995–)，男，四川眉山人，硕士生，研究方向为SAR图像目标检测识别、PolSAR图像地物分类。E-mail: xdluestc@outlook.com
刘畅：刘 ? 畅(1978–)，男，山东烟台人，研究员，博士生导师。2006年在中国科学院电子学研究所获得博士学位，现担任中国科学院电子学研究所研究员、博士生导师。主要研究方向为SAR系统及其相关SAR成像处理技术等。E-mail: cliu@mail.ie.ac.cn
通讯作者:刘畅 cliu@mail.ie.ac.cn
中图分类号:TN958计量
文章访问数:1891
HTML全文浏览量:625
PDF下载量:217
被引次数:0
出版历程
收稿日期:2019-03-04
修回日期:2019-03-19
网络出版日期:2019-05-05
PolSAR Terrain Classification Based on Fine-tuned Dilated Group-cross Convolution Neural Network
XIAO Dongling,LIU Chang,
Institute of Electronics, Chinese Academy of Sciences, Beijing 100190, China
University of Chinese Academy of Sciences, Beijing 100049, China
Funds:The National Natural Science Foundation of China (61471340), The State Key Research Development Program (2017YFB0503001)
More Information
Corresponding author:LIU Chang, cliu@mail.ie.ac.cn
摘要
摘要:在极化合成孔径雷达(PolSAR)地物分类研究中，基于实数CNN的分类算法无法充分利用PolSAR图像丰富的通道相位信息，并且在逐像素预测中存在大量冗余计算，导致分类效率较低。针对以上问题，该文采用一种复数域下的像素映射深度模型，实现低采样率下精确且高效的PolSAR地物分类。为充分使用PolSAR数据的通道相位信息，该文基于一种编组-交叉卷积网络(GC-CNN)将分类模型推广到复数域，并利用网络提取的复数特征及其对应的相位和幅度来实现更精确的分类；为加快分类速度，该文还采用了一种精调的膨胀编组-交叉卷积网络(FDGC-CNN)来实现像素到像素的直接映射，并进一步提升了分类精度。在基于AIRSAR平台的16类地物数据和E-SAR平台的4类地物数据的实验中，该文采用的FDGC-CNN模型相较于SVM分类器和实数CNN模型，能够更准确和更高效地实现多类别地物分类，全局分类精度分别为96.94%和90.07%、总耗时4.22 s和4.02 s。
关键词:极化合成孔径雷达(PolSAR)/
地物分类/
膨胀卷积/
编组-交叉卷积
Abstract:In the study of the terrain classification based on the Polarimetric Synthetic Aperture Radar (PolSAR), the algorithms based on general CNN do not fully utilize the phase information in different channels, and the pixel-by-pixel classification strategy with extensive redundant computation is inefficient. To mitigate these problems, a deep pixel-to-pixel mapping model in the complex domain is used for achieving a fast and accurate PolSAR terrain classification at a low sampling rate. To completely utilize the phase information, this study uses Group-Cross CNN to extend the original model to the complex domain allowing complex-number input signals and significantly improving the classification accuracy. In addition, to speed up the algorithm, a Fine-tuned Dilated Group-Cross CNN (FDGC-CNN) was adopted to directly achieve pixel-to-pixel mapping as well as improve accuracy. We verified the adopted model on two PolSAR images comprising 16 classes terrains from the AIRSAR and 4 classes terrains from the E-SAR. According to our model, the overall classification accuracies were 96.94% and 90.07% respectively while the running time was 4.22 s and 4.02 s respectively. Therefore, FDGC-CNN achieved better accuracy with higher efficiency compared to SVM and traditional CNN.
Key words:PolSAR/
Terrain classification/
Dilated convolution/
Group-cross convolution
PDF全文下载地址:
https://plugin.sowise.cn/viewpdf/198_59c4457a-f4b2-4b80-91b5-d69e1d912b5c_R19039
