黄河流域内陆地表水体提取方法研究

Extraction Method of Spectral Information of Inland Surface Water Body in Yellow River Basin

[目的]对黄河流域内陆地表水体提取方法进行对比分析,为有效提取含沙量大的黄河流域水体提供方法指导。[方法]采用目前提取水体效果较好的两种方法——改进的水体指数法(MNDWI)和线性光谱混合模型(LSMM),以Landsat 8OLI数据为例,选择黄河流域水库、湿地、湖泊和河流作为研究对象,将其划分为2大类,即水体和非水体,利用高分辨率影像进行精度分析,研究两种方法的区域适应性。[结果](1)利用线性光谱混合模型在提取水库、湿地和湖泊比改进的水体指数模型精度更高;(2)利用这2种方法在提取面积较大、分布集中的水体比提取细长型分布的线状河流效果更好。[结论]混合像元在高分辨率的影像中也是存在的,在水体提取的时候,利用线性光谱混合模型考虑了混合像元对水体提取的影响,极大提高了精度,试验证明线性光谱混合模型优于改进的水体指数法。

[Objective] Extraction methods of spectral information of inland surface water in the Yellow River basin were elucidated and compared to provide guidance for the extraction of spectral information of water bodies with large sediment concentration in the Yellow River basin. [Methods] Two methods that were thought effective at present were chose to extract the spectral information of water body. They were modified normalized difference water index（MNDWI） and linear spectral mixture model. Landsat 80LI imagery of reservoir, wetland, lake and river in the Yellow River basin was exemplified to analyze the accuracies of the two methods, and to discuss the regional applicability. In which, the study area was divided into two categories： water and non-water, and high resolution imagery was referred. [Results] The accuracy of linear spectral mixture model in extracting spectral information of reservoir, wetland and lake was higher than that of the MNDWI. The two methods performed better in the large-area water bodies, such as lakes and reservoirs, than in the linear like body as rivers. [Conclusion] In high resolution image, mixed pixels were also existed. Based on that, the linear spectral mixture model had covered the effect of mixed pixels on the spectral information extraction from water bodies, whereby it remarkably improved the extraction precision. The linear spectral mixture model is superior to the modified normalized difference water index.