基于FY3A/MERSI影像的几种常用水体提取方法的比较分析

Comparative analysis of several commonly used water extraction method based on FY3A/MERSI imagery

随着遥感技术的飞速发展,利用遥感数据来进行水资源的监测、调查和分析已经成为一种必然的趋势。通过选取新疆博湖县境内中国最大的内陆淡水湖-博斯腾湖为研究区,FY3A/MERSI影像为数据源,利用监督分类法从Landsat-ETM+影像提取水体,提取结果作为FY3A/MERSI影像水体提取精度验证的底图。采用单波段阈值法、基于阈值的多波段谱间关系法和基于阈值的水体指数法从FY3A/MERSI影像提取研究区水体,基于混淆矩阵法,提取结果分别与Landsat-ETM+影像底图作对比分析,最终得出结论,基于阈值的水体指数法中的归一化差异水体指数法提取研究区水体的总体精度最高,为96.37%,Khat为0.915。

With the development of remote sensing technology,monitoring,investigation and analysis of water resources using remote sensing data has become an inevitable trend. In this paper,Lake Bosten at Bohu County in Xinjiang,the largest inland freshwater lake in China,is selected as the study area. taking FY3A/MERSI images as the data source,extracting water body from Landsat-ETM＋ image with supervised classification method,and the extracting results are used as the base map for verifying the accuracy of water data extracting from FY3A/MERSI images. The one band threshold methods,the threshold-based multi-band relationship methods and the threshold-based water body index method are used for extracting water data of the study area from FY3A/MERSI images. Based confusion matrix method,the extracting results were compared with Landsat-ETM ＋ images,which indicated that the normalized difference water index method of the threshold-based water body index methods has the highest overall accuracy of 98.82% and Khat 0.960.