城市地表水是城市生态环境的重要组成部分, 地表水环境高光谱遥感是高光谱遥感的重要应用方向, 水体提取是地表水环境高光谱遥感的第一步, 其主要任务是从高光谱遥感数据中提取地表水水体轮廓。 基于光谱指数的水体提取方法充分利用光谱信息, 计算简单, 实现容易, 提取效果优异。 归一化植被指数(NDVI)、 归一化水体指数(NDWI)、 高光谱差异化水体指数(HDWI)和基于指数的水体指数(IWI)等光谱指数已经广泛应用于湖泊、 大江大河等开阔水体提取。 近些年来, 随着成像光谱技术的发展, 高光谱遥感数据的获取能力也突飞猛进, 空间分辨率和光谱分辨率不断提高。 与江河湖基本在流域内沿地形分布不同, 城市地表水一般细小, 纵横交错, 形成河网。 在高光谱遥感数据用于城市体表水提取时, 其面临的图像空间分辨率、 地物类型和地物复杂等, 与江河湖水体提取有很大不同。 因此, 需要对这些常用的光谱指数在城市地表水提取中的适宜性进行评价。 以此做为出发点和目标, 以河网密布的江南水乡中国浙江省嘉兴市为研究对象, 以应用型航空成像光谱仪(Airborne imaging spectrometer for applications, AISA)获取的高空间分辨率机载高光谱遥感数据为数据源, 通过Youden指数确定最佳阈值, 将总体分类精度、 错分误差、 漏分误差、 Kappa系数作为衡量指标, 分析评价了NDVI, NDWI, HDWI和IWI 4种光谱指数在城市河网提取中的适宜性。 结果表明, 阴影与水体光谱变化趋势类似, 是造成水体提取过程中高错分误差的主要因素。 四种指数都可以准确抑制落在植被中的阴影, 但无法有效抑制落在建筑物中的阴影。 HDWI虽然可以在一定程度上抑制建筑物中的阴影, 但是无法有效地抑制亮建筑物背景。 通过对不同类型水体和阴影(笼罩下地物)光谱的进一步分析, 虽然水体和阴影光谱曲线变化趋势相似, 均在560～600 nm附近存在波峰, 但是水体和阴影波峰高度存在差异, 水体波峰值较大而阴影波峰值较低。 因此, 通过充分挖掘水体和阴影在560～600 nm处光谱反射信息, 有望进一步抑制建筑物阴影, 提高城市河网水体提取精度。

Urban surface water is an important part of urban ecological environment. Hyperspectral remote sensing of surface water environment is an important application direction of hyperspectral remote sensing. Water extraction is the first step of hyperspectral remote sensing of surface water environment. Its main task is to obtain the contour of the surface water body from hyperspectral remote sensing data. The water body spectral index makes full use of the spectral information, and the calculation is simple, the implementation is easy, and the extraction effect is excellent. Spectral indices such as normalized difference vegetation index (NDVI), normalized difference water index (NDWI), hyperspectral difference water index (HDWI) and index of water index (IWI) have been widely used in the extraction of open water bodies such as lakes and large rivers. In recent years, with the development of imaging spectroscopy technology, the acquisition capability of hyperspectral remote sensing data has also advanced rapidly, and spatial resolution and spectral resolution have been continuously improved. The rivers and lakes are basically distributed along the topography in the basin while the urban surface water is generally small, criss-crossed, forming a river network. When hyperspectral remote sensing data are used for urban surface water extraction, the spatial resolution of the image, the type of features and the complexity of the ground objects are very different from those of rivers and lakes. Therefore, the applicability of these commonly used spectral indices in urban surface water extraction needs to be evaluated. This article is based on this starting point and goal, taking the Jiaxing City, Zhejiang Province in China, which is in Jiangnan Water Town and has a dense river network as the research object, and using the high spatial resolution airborne hyperspectral remote sensing data acquired by airborne imaging spectrometer for applications (AISA) as data source. The optimal threshold is determined by Youden index. The overall accuracy, commission error, omission error and Kappa coefficient are used as the accuracy evaluation indicators. The suitability of NDVI, NDWI, HDWI and IWI in urban river network extraction was analyzed and evaluated. The results show that the trend of the shadow spectrum is similar to the water spectrum, and is the main factor causing high commission errors in the water body extraction. All four indices accurately suppress the shadows that fall in the vegetation, but do not effectively suppress the shadows that fall in the buildings. Although HDWI can suppress shadows cast in buildings to a certain extent, it cannot effectively suppress the bright buildings. Through further analysis of the spectrum of different types of water bodies and (the ground objects under) shadows, the water and shadow spectral curves are similar, and there are peaks around 560～600 nm, but the heights of water and shadow peaks are different. The water wave peaks are larger while the peak value of the shadow wave is lower. Therefore, by fully excavating the spectrum reflectance information at 560～600 nm in water bodies and shadows, it is expected to further suppress building shadows and improve the accuracy of water extraction in urban river networks.