Abstract: This paper explored a hybrid approach for small linear river-water extraction by combining spatial information, spectral computation and global & local iterative algorithm. The process comprises four steps. Firstly, the river distribution maps which can be served as prior knowledge were derived from DEM data. Then a specific target area was identified using spatial analysis technique associated with river distribution mentioned above, furtherly we can get the preliminary water distribution information after applying water spectral index computation and global threshold segmentation. Finally, the precise extraction of river water can be achieved by the combination of local water index analysis, self-adaptive threshold selection and iterative computation. In this study, we adopted this approach to extract river-water information with ETM+ data covering the middle reaches of YILI river, the results show that the proposed method can accomplish river network mapping effectively and precisely. Furthermore, it increased the direction of information extraction and initiative of computation by minimizing the influence of background spectral interference to the extreme extent.

Key words: River water, Auto-extraction, Adaptive, Iterative computation