Research on Urban Water Body Extraction Using Knowledge-based Decision Tree

In view of the spectral mixing between water body,building shadow,asphalt road and dense vegetation in urban environment,a knowledge-based decision tree combining spectral and spatial features is constructed to extract water body thematic information in this paper.Firstly,dark objects in urban environment are extracted using threshold of reflectance in SWIR.Secondly,dense vegetation and asphalt road are eliminated according to their reflectance in NIR and R respectively.Thirdly,differences in spatial density are used to eliminate building shadow.Finally,area threshold is used for supplementary recognition of water body.The consideration of dark objects in urban water body extraction,and the using of spatial density described by DBSCAN in discriminating water body from building shadow are two main differences between the proposed decision tree and state-of-art methods.SPOT-5 multispectral imagery of Beijing is used to validate the proposed knowledge-based decision tree.The detection rate is 86.18% and false alarm rate is 13.82%.It can be concluded that the proposed model is an effective method in water body thematic information extraction based on medium-resolution multi-spectral imagery in urban environment.