陈嘉琪^1,,,
刘祥梅¹,
李宁^{2, 3},
张燕¹
1.河海大学计算机与信息学院 南京 211100
2.河南大学计算机与信息工程学院 开封 475004
3.河南省大数据分析与处理重点实验室 开封 475004
基金项目:国家自然科学基金(61771183, 61601437)，中央高校基础研究基金(2016B07114)，河南省科技攻关计划项目(192102210082)，河南省青年人才托举工程(2019HYTP006)，中国博士后科学基金(2013M541035)

详细信息

作者简介:陈嘉琪：男，1984年生，副教授，硕士生导师，研究方向为雷达信号处理及应用，计算电磁学
刘祥梅：男，1997年生，硕士生，研究方向为SAR图像处理
李宁：男，1987年生，教授，博士生导师，研究方向为SAR信号与图像处理
张燕：女，1994年生，硕士生，研究方向为遥感图像超分辨率重建

通讯作者:陈嘉琪　cjq19840130@163.com

中图分类号:TN959.1

计量

文章访问数:511
HTML全文浏览量:150
PDF下载量:79
被引次数:0

出版历程

收稿日期:2020-05-08
修回日期:2020-12-05
网络出版日期:2020-12-17
刊出日期:2021-03-22

A High-precision Water Segmentation Algorithm for SAR Image and its Application

Jiaqi CHEN^1,,,
Xiangmei LIU¹,
Ning LI^{2, 3},
Yan ZHANG¹
1. College of Computer and Information, Hohai University, Nanjing 211100, China
2. School of Computer and Information Engineering, Henan University, Kaifeng 475004, China
3. Henan Key Laboratory of Big Data Analysis and Processing, Henan University, Kaifeng 475004, China
Funds:The National Natural Science Foundation of China (61771183, 61601437), The Fundamental Research Funds for the Central University (2016B07114), The Plan of Science and Technology of Henan Province (192102210082), The Youth Talent Lifting Project of Henan Province (2019HYTP006), China Postdoctoral Science Foundation (2013M541035)


摘要
摘要:合成孔径雷达(SAR)图像水域分割在湖泊、河流等陆地水文监测领域有重要的研究意义。由于SAR图像分辨率不足所导致的陆地与水域边界模糊, 会影响水域分割精度。该文以中国青藏高原地区的多庆错湖为研究对象，使用Sentinel-1A SAR图像数据，综合运用深度残差模型、通道注意力与亚像素卷积，提出一种基于亚像素卷积的增强型通道注意力深度残差超分辨网络，对滤波后的SAR图像进行重建、水域轮廓提取与精度分析。通过比较不同超分辨算法下的重建结果及水域轮廓提取精度，该文算法在重建效果与提取精度上都较传统方法有明显提升，并具有很好的鲁棒性。
关键词:合成孔径雷达/
超分辨率重建/
水域分割
Abstract:Water segmentation of Synthetic Aperture Radar (SAR) is of great significance in land hydrological monitoring, such as lakes and rivers. Water segmentation accuracy is influenced by the blurring of the boundary between land and water region because of the insufficient resolution of SAR image. Sentinel-1A SAR image is used to study the Duoqing Co in the Tibetan Plateau of China. This paper integrates the enhanced deep residual block, channel attention mechanism and sub-pixel convolution, an enhanced channel attention deep residual network is proposed based on sub-pixel to reconstruct the filtered SAR image, extract the water contour and analyze the accuracy. By comparing the reconstruction results of different super-resolution algorithms and the accuracy of water contour extraction, this algorithm, with great robustness, is obviously better than the traditional method in both reconstruction effect and extraction accuracy.
Key words:Synthetic Aperture Radar (SAR)/
Super-resolution reconstruction/
Water segmentation



PDF全文下载地址:

https://jeit.ac.cn/article/exportPdf?id=f4ea4b4c-2ca8-4edb-9884-9f2ad867467b