摘要
应用遥感技术提取水体信息已成为研究热点之一。我国卫星高分三号(GF-3)雷达数据具有全天时全天候工作和高空间分辨率的特点,但是容易受到斑点噪声的影响,而且包含的对象信息非常有限。多光谱图像包含丰富的光谱信息,可以填补雷达数据的信息缺陷,但是空间分辨率相对较低。为了提高遥感图像中水体的分类精度,采用多源数据融合办法,首次提出GF-3雷达图像与Sentinel-2A多光谱图像融合算法执行水体提取任务。该算法使用Gram-Schmidt变换进行融合,采用随机森林实现水体的最终识别和提取。实验结果表明,该算法在水体提取任务中比基于单一的雷达图像或多光谱图像方法具有明显优势,获取水体的识别精度显著提高。
The application of remote sensing technology to extract water information has become one of the research focuses.GF-3,a synthetic aperture radar(SAR)satellite from China,can work all day and all weather,and has high spatial resolution.However,it is vulnerable to speckle noise and has very limited object information.Multi-spectral images contain abundant spectral information,and can be complementary to radar data,but its spatial resolution is relatively low.In order to improve the classification accuracy of water bodies in remote sensing images,the multi-source data fusion method is adopted,and the fusion algorithm of GF-3 SAR image and Sentinel-2A multispectral image is proposed for the first time to perform the water extraction task.In the method,Gram-Schmidt transform is used for fusion,and the random forest is applied to realize the final recognition and extraction of water.The experimental results show that this algorithm has obvious advantages over the single SAR image or multispectral image method in extraction task,and the recognition accuracy of the obtained water body is significantly improved.
作者
童莹萍
冯伟
全英汇
倪卓娅
邢孟道
TONG Yingping;FENG Wei;QUAN Yinghui;NI Zhuoya;XING Mengdao(School of Electronic Engineering,Xidian University,Xi’an 710071,China;National Satellite Meteorological Center,Beijing 100081,China;Academy of Advanced Interdisciplinary Research,Xidian University,Xi’an 710071,China)
出处
《无线电工程》
北大核心
2021年第8期750-753,共4页
Radio Engineering
基金
国家自然科学基金资助项目(61772397,12005169)
国家重点研发计划资助项目(2016YFE0200400)
数字地球重点实验室开放基金项目(2019LDE005)
陕西省科技创新团队(2019TD-002)
中央高校基本业务费(XJS200205)。
关键词
水体提取
图像融合
分类
随机森林
遥感
water extraction
image fusion
classification
random forest
remote sensing
