摘要
水体信息的高精度提取是区域水资源利用研究的基础和关键。为改善平原河网区水体遥感信息提取中城镇、阴影和裸土等像元混淆问题,基于多时相高分一号影像,提出时序归一化水体指数(NDWI)、归一化植被指数(NDVI)和近红外波段反照率信息相结合的决策树水体信息提取方法,并以江淮下游典型平原区开展实证研究。试验结果表明,多时相光谱信息的使用有效改善了阴影和裸土等像元的混淆,提高了较小水体的提取精度,总体精度为93%,Kappa系数为0.85,更合理地反映了水面率呈自西向东逐渐减少的空间特征,研究成果可为其他平原区水体信息提取提供借鉴。
High-accuracy extraction of water body information is the base of and the key to the study of the regional water resources utilization.In order to solve the mixture of pixels such as the towns,shadow and bare earth in the extraction of remote sensing information of water body in Plain River Network Area,based onmulti-temporal GF-1 image,this paper proposes a decision tree model combining the Normalized Differential Vegetation Index(NDVI),Normalized Differential Water Index(NDWI),and near-infrared albedo for water body information extraction,and conducts a study taking the typical plain atthe lower reaches of Yangtze and Huai River.The results show that compared with other methods including single-band threshold and Shadow Water Index(SWI),the multi-temporalspectral information in newly proposed method can effectively eliminate the mixture of pixels such as the towns,shadow and the bare earth,improve the extraction accuracy of smaller water bodies with an overall accuracy of 93%and a Kappa coefficient of 0.85,and reasonably show a spatial distribution of water surface rate with a declining trend from West to East,which could provide a reference for the information extraction of water body in other plains as well.
作者
廖钰冰
周峰
LIAO Yubing;ZHOU Feng(School of Civil Engineering,Yancheng Institute of Technology,Yancheng 224051,China;School of Environmental Science and Engineering,Yancheng Institute of Technology,Yancheng 224051,China)
出处
《人民珠江》
2020年第1期34-38,121,共6页
Pearl River
基金
国家自然科学基金(41401035)
关键词
高分一号
水体信息提取
决策树
平原河网区
GF-1
extraction of water body information
decision tree
plain river network area
