刘宏洁,宋文龙,刘昌军,卢奕竹,曲伟,唐锐,桂荣洁.基于归一化水体指数及其阈值自适应算法的水体遥感反演效果分析[J].中国水利水电科学研究院学报,2022,20(3):251-261
基于归一化水体指数及其阈值自适应算法的水体遥感反演效果分析
Analysis of water remote sensing inversion effect of Normalized Difference Water Index and its adaptive threshold correction algorithm
投稿时间:2021-10-20  
DOI:10.13244/j.cnki.jiwhr.20210264
中文关键词:  水体遥感反演  归一化水体指数  均值漂移聚类算法  最大类间方差算法
英文关键词:water body remote sensing inversion  Normalized Difference Water Index  mean shift clustering algorithm  maximum variance between classes algorithm
基金项目:国家重点研发计划项目(2018YFC1508702);江苏省水利科技项目(2021081);辽宁省"兴辽英才计划"项目(XLYC2002007)
作者单位E-mail
刘宏洁 中国水利水电科学研究院, 北京 100038  
宋文龙 中国水利水电科学研究院, 北京 100038 songwl@iwhr.com 
刘昌军 中国水利水电科学研究院, 北京 100038  
卢奕竹 中国水利水电科学研究院, 北京 100038  
曲伟 中国水利水电科学研究院, 北京 100038  
唐锐 陕西航天技术应用研究院, 陕西 西安 710000  
桂荣洁 中国水利水电科学研究院, 北京 100038
首都师范大学 资源环境与旅游学院, 北京 100871 
 
摘要点击次数: 514
全文下载次数: 802
中文摘要:
      归一化水体指数(NDWI)是水体遥感反演的一种重要指标,其阈值及修正直接影响反演结果的精度。基于Landsat-8与GF-1光学多光谱影像,使用归一化水体指数法阈值0(TH0)、最大类间方差法(OTSU)自适应阈值(THotsu)与均值漂移聚类算法(Mean-Shift)自适应阈值(THMS)分别对典型正常水体、云雾覆盖水体、富营养化水体、高含泥沙水体进行水体遥感提取与效果分析,结果表明:正常水体以TH0为阈值提取精度最高,THMS提取精度次之,THotsu提取精度最差;而云雾覆盖水体、富营养化水体以及含泥沙水体使用THMS提取精度最高,尤其少量云雾覆盖下的水体,THMS具有更明显的优势,TH0提取精度次之,THotsu提取精度最差;对于不同的阈值,Land?sat-8比GF-1总体表现出更高的水体提取精度。Mean-Shift算法应用于NDWI阈值修正与水体遥感反演具有快速、水质适应性强、效果稳定的优势,对尤其是复杂条件下的水体信息遥感反演具有较好的提取效果。
英文摘要:
      Normalized Difference Water Index (NDWI) is an important index of water remote sensing inversion. Its threshold and correction directly affect the accuracy of inversion results. Based on landSAT-8 and GF-1 optical multispectral images, Normalized water index method threshold 0 (TH0), maximum interclass variance method adaptive threshold (THotsu) and mean-shift clustering algorithm adaptive threshold (THMS) were used to extract typical normal water body, cloud covered water body, eutrophication water body and water body with high sediment content. The results show that:The extraction accuracy of normal water body was the highest with TH0 as the threshold, followed by THMS and THotsu. The extraction accuracy of THMS was the highest in water covered by cloud, eutrophication and sediment, especially in water covered by a small amount of cloud, THMS had a more obvious advantage, TH0 was the next, and THotsu was the worst. For different thresholds, Landsat-8 showed higher water extraction accuracy than GF-1. Mean-Shift applied to NDWI threshold correction and water remote sensing inversion has the advantages of fast, strong adaptability to water quality and stable effect, and has good extraction effect for water information remote sensing inversion under complex conditions.
查看全文  查看/发表评论  下载PDF阅读器
关闭