基于高光谱遥感数据的城市河网水体提取方法比较

Comparison of Water-body Extraction Methods in Urban River Network Based on Hyperspectral Data

目前水体提取多局限于基于多光谱遥感数据的研究,高光谱遥感的出现为水体精细光谱获取和水体提取提供了可能.以机载高光谱遥感数据源为基础,构建了适用于可见-近红外高光谱数据的水体提取决策树,充分利用高光谱数据的光谱丰富性,设定了决策树的最优波段和提取阈值.以2012年AISA成像光谱获取的嘉兴城区高光谱数据作为验证数据,采用检出率和误检率对所提出的水体提取算法进行验证和评价.实验结果表明,对比单波段阈值法、谱间关系法、归一化水体指数法、分类法和决策树这5种方法,决策树水体提取方法效果最好且误检率最低.

Water body extraction methods are mainly based on multispectral remote sensing data recently. The emergence of hyperspectral remote sensing makes it possible to acquire fine spectra of water and extract water body with high precision. Based on airborne hyperspectral remote sensing images as data source,water body extraction decision-tree suitable for VNIR hyperspectral data is proposed, taking full advantage of spectral richness of hyperspectral remote sensing data, the best bands and extraction threshold are determined. Hyperspectral imaging data of Jiaxing city acquired by AISA in 2012 is used as the verification data, the detection rate and false alarm rate are used to evaluate water body extraction results. Decision-tree method, single-band threshold method, multiband threshold method （spectrum-hotometric method,NDWI） and sub-categories method are applied to extract water body. The results show that decision tree method is the best with higher detection rate and lower false alarm rate.