关键词: 水体提取, 归一化差分水体指数, 支持向量机, 改进光谱角匹配

Abstract: Quickly and accurately obtaining water information from satellite remote sensing imagery has become an important technical means in the field of water resources investigation and monitoring,wetland protection,and flood disaster assessment. In this paper,a new water extraction method of a modified spectral angle mapping (MSAM) is proposed and GF-4's PMS images are used as research data.Two typical study areas are investigated, which are selected from the areas of middle-east of Inner Mongolia and middle-lower reaches of the Yangtze River.To evaluate the efficacy of the proposed MSAM method,four compared methods including single band threshold,NDWI threshold, support vector machine (SVM) and spectral angle mapping (SAM) are used in water extraction.Then accuracy assessments are taken on the water extraction results of the five different methods.The results show that the proposed method can not only extract water information accurately,but also distinguish water and the cloud shadows well and have a good ability in extracting small water bodies.And it obtains the highest accuracy compared with the other four water extraction methods,with the water extracting accuracies of the two study areas reaching 99.86% and 98.37% respectively.To sum up,the proposed method can be effectively used to extract water information of GF-4 remote sensing imagery.

Key words: water extraction, normalized difference water index, support vector machine, modified spectral angle mapping