基于GF-1卫星影像的改进SWI水体提取方法

Method of water information extraction by improved SWI based on GF-1 satellite image

大尺度高精度山区河流信息提取是我国干旱区水资源开发利用的关键技术,而利用遥感影像提取水资源信息存在水体与山区阴影难以区分的瓶颈。以GF-1号卫星2 m分辨率全色波段影像和8 m分辨率多光谱影像为数据源,选取新疆特克斯河流域巴喀勒克水库为研究区,提出改进的阴影水体指数法(modified shade water index,MSWI)进行水体信息提取;同时运用单波段阈值法、NDWI法、单波段法与阴影水体指数法(shade water indes,SWI)相结合的决策树分类法(简称SWI)以及单波段法与MSWI相结合的决策树分类法(简称MSWI)分别对研究区水体信息进行提取,并进行了对比分析。研究结果表明,前2种方法与SWI和MSWI法相比,效果稍差;而SWI和MSWI法分类效果较好,其中MSWI比SWI法分类总精度高0.94%,提高了高分辨率遥感影像的解译精度,可为国产高分系列卫星影像在干旱区水资源信息提取中的应用提供技术支持。

High - precision information extraction of mountainous rivers is a key technology for development and utilization of water resources in arid areas of China. Nevertheless, the utilization of remote sensing images cannot distinguish water form mountain shadows. In this paper, the authors used GF - 1 satellite images with resolution of 2 m and 8 m as the data source, selected Baka Luck reservoirs as the study area, and put forward an improved method （ modified shadow water index, MSWI） for water information extraction. At the same time, the authors used the single- band threshold method, the NDWI method, the single band method combined with the SWI decision tree classification（SWI） and the single band method combined with the MSWI decision tree classification （MSWI） respectively to extract water information in the study area. The results show that, compared with the SWI and the MSWI method, the first two methods have relatively poor performance. The SWI and MSWI classification effect is good and the total classification accuracy of ＇MSWI is increased by 1.22% relative to the SWI method. It can provide technical support for the domestic high series satellite image information extraction in water resources in arid regions.