一种基于城市水体指数与分形几何算法的OLI遥感影像水体提取方法

针对严重污染的城市水体与道路、建筑物、阴影等易于混分，以及遥感水体提取结果不连续、存在斑点问题，本文以广州市流溪河与东江水系为研究对象，基于2016年与2017年OLI遥感影像，采用本文新提出的城市水体指数法（CWI），同时结合分形几何算法，通过设置形状面积等特征，实现城市复杂环境下的水体信息的自动提取。并与单通道算法、改进的归一化差异水体指数（MNDWI）算法、支持向量机法（SVM）与光谱角度法的水体提取结果进行对比分析。结果表明：SVM算法出现大量斑点，其次为MNDWI水体指数算法，光谱角度算法与单通道算法斑点较少，但水体提取结果不连续，部分河道漏分。本文提出的算法能够克服山体阴影、道路、建筑物等影响，实现城市污染水体以及一般水体连续、准确提取。本文的提取结果可为水资源调查、洪水灾害预测评估、水利规划、环境监测等工作提供基础数据支撑。

Extracting Water Information from OLl Remote Sensing Images Based on City Water Index and Fractal Geometry

In the remote sensing images,roads,buildings and shadows are difficult to distinguish from severely polluted urban water bodies.And the extracting results from remote sensing images are not continuous and exists some spots.Aimed at these problems,this paper based on the 2016 and 2017 OLI remote sensing images,city water index (CWI) method is used.The fractal geometry algorithm and the shape area are used to automatic extract the water information in urban complex environment.The results are compared with single channel algorithm,modified normal difference water index (MNDWI) algorithm,support vector machine method (SVM) and spectral angle method.The results shows that there are a large number of spots in the SVM algorithm,followed by the MNDWI water body index algorithm.The spectrum angle algorithm and the single-channel algorithm are with fewer spots,but the water extraction results are discontinuous and part of the channel leakage.The algorithm proposed in this paper can continuously and accurately extract urban water bodies,and can also overcome the influence of mountain shadows,roads and buildings.Compared with other methods,this method is greatly improved.The results can provide basic data support for water resources investigation,flood disaster prediction and assessment,water conservancy planning and environmental monitoring.