资源科学 ›› 2019, Vol. 41 ›› Issue (4): 790-802.doi: 10.18402/resci.2019.04.15

• 水资源 • 上一篇    下一篇

一种稳定阈值的湖泊水体信息提取方法

朱小强1,2,3, 丁建丽1,2, 夏楠1,2, 郭家新1,2, 张淑霞1,2, 杨童童1,2, 王敬哲1,2, 李晓航1,2   

  1. 1. 新疆大学资源与环境科学学院,乌鲁木齐 830046;
    2. 新疆大学绿洲生态教育部重点实验室,乌鲁木齐 830046;
    3. 中电科新型智慧城市研究院有限公司,深圳 518035
  • 收稿日期:2017-04-27 修回日期:2017-08-23 发布日期:2019-04-25
  • 通讯作者: 国家自然科学基金项目(41771470); 2016新疆研究生科研创新项目(XJGRI2016013); 国家自然科学基金地区项目(41561089)
  • 作者简介:朱小强,男,新疆奎屯人,硕士研究生,主要从事干旱区资源与环境遥感应用研究。E-mail: z_xiao_qiang@163.com

Temperature vegetation water index: A novel stabilized threshold method for lake surface water mapping

ZHU Xiaoqiang1,2,3, DING Jianli1,2, XIA Nan1,2, GUO Jiaxin1,2, ZHANG Shuxia1,2, YANG Tongtong1,2, WANG Jingzhe1,2, LI Xiaohang1,2   

  1. 1. College of Resources and Environment Science, Xinjiang University, Urumqi 830046, China;
    2. Key Laboratory of Oasis Ecology of Ministry of Education, Xinjiang University, Urumqi 830046, China;
    3. The Smart City Research Institute of China Electronics Technology Group Corporation, Shenzhen 518035, China
  • Received:2017-04-27 Revised:2017-08-23 Published:2019-04-25

摘要: 针对提取水体信息时阈值设定模糊问题,提出一种具有稳定阈值的在反射率为0时水体与其他地物分离度最好的自动水体信息提取方法。利用不同时间段的艾比湖、博斯腾湖、鄱阳湖、阿拉湖的Landsat 8卫星影像为数据源,通过分析水体与不同地物的反射率及主分量特征,构建一种新型水体指数——温度植被水体指数(Temperature Vegetation Water Index, TVWI),利用监督分类得到的水体面积为参考值,分别与传统水体提取方法进行对比验证。结果表明:①利用TVWI分别对2013年与2016年的艾比湖进行水体信息提取,总体精度分别为98.51%、97.33%,博斯腾湖为99.66%,鄱阳湖为98.06%,阿拉湖为99.72%,总体精度较高;②在TVWI指数中,水体与非水体区域区分度较高,在水体边界处的非水体区域出现极高的噪音值,而水体区域值大于0;③通过普适性分析得知,TVWI对湖泊适应性较好,对不同类型的湖泊水体提取精度均较高。因此,利用TVWI进行高精度无模糊阈值设定的水体信息提取是可行的。

关键词: 水体提取, 阈值, 温度植被水体指数, 噪音, 艾比湖, 博斯腾湖, 鄱阳湖, 阿拉湖

Abstract: Considering that the traditional water indices have no exact threshold value to distinguish water body from other land cover types, in this study a novel method was proposed, in which the reflectance value “0” is used as a segmentation line to distinguish water body from other land cover types. Landsat 8 satellite data of Ebinur Lake, Bosten Lake, Poyang Lake, and Alakol Lake from different years were acquired and used to analyze the reflectance value and spectral and principal component space features of water and other land cover types. A novel water index named temperature vegetation water index (TVWI) was constructed. Maximum likelihood classification was used to extract water areas as the ground truth data. Compared with traditional water extraction indices, the results indicate that the overall TVWI water extraction accuracy in May 2013 and May 2016 is 98.51% and 97.33%, and Kappa coefficients are 0.97 and 0.95 correspondingly in Ebinur Lake. The overall TVWI water extraction accuracy in Bosten Lake is 99.66%, Poyang Lake is 98.06%, and Alakol Lake is 99.72%. The overall accuracy is better than the traditional methods, which implies that extracting water information by using TVWI method is feasible. In TVWI the reflectance value of 0 is the segmentation value to classify water and non-water areas. There are higher noise values at the lakefront but the values of water area are higher than 0, so there is a visible distinction between water and other land cover types. As a result, water information can be easily extracted by using the TVWI method. In conclusion, extracting Ebinur water information using the Landsat 8 satellite data from 29 May 2013 and 21 May 2016 yielded better result, and the overall accuracy and Kappa coefficient are the best compared with the three traditional water indices that we compared. The adaptation of TVWI to different lakes was analyzed and the accuracy of water extraction in different types of lake is higher than other methods.

Key words: water information extraction, threshold, Temperature Vegetation Water Index (TVWI), noise value, Ebinur Lake, Bosten Lake, Poyang Lake, Alakol Lake