基于TM影像的珠江三角洲河道快速提取方法研究

Fast Extraction Technique of River Channels Based on Landsat TM

快速、准确提取河流水体信息是河道开发和治理的基础。在综合分析前人已有研究成果的基础上,以珠江三角洲佛山市境内河网为例,通过对TM影像上水体的遥感信息机理分析,建立多种水体指数模型,并对这些指数模型进行对比分析,结果发现比值指数模型最适合于网河区河流水体的提取。在此基础上,对比值指数模型提取结果进行空间聚类得到网状水系图,并利用形状指数进行修正,将遗漏的细小河流补充进来,最后得到河道水体专题图,精度评价的结果显示其精度可以达到93.7%,说明该方法具有较好的可用性和有效性。特别是应用于河网密布、鱼塘众多,同时人为干扰强烈区域的河道信息提取时具有重要的现实与理论意义。

It is the basis for river management, development and remediation to extract and acquire accurate water information using remote sensing and GIS techniques. A net river region of Foshan in the Pearl River Delta was selected as study area. After analyzing the spectral characteristics of river water bodies and other objects in Landsat TM images, several water indices models were established. Among these models, ratio water indices model（RWI） proved to be the best one. Then, the results of RWI were taken to get water thematic images by using special clustering analysis method, as a result, the main water information were extracted from other spectrums. In addition to these, the missing small rivers were checked out by the shape index. Accurate assessment indicated that this technique is feasible and effective, furthermore, accuracy can reach 93.7%. That is, the method has important academic and practical significance,especial for the water information of net river region, where the river web is more complicated and human interference is intense.