遥感图像非监督分类分析江宁县江滩钉螺孳生地植被特征

目的：从遥感图像中提取江宁县江潍钉螺孳生地的植被特征。方法：在ERDAS IMAGINE8．5软件支持下，根据LANDSAT7 ETM图像中各波段的光谱特征，选择对植被敏感的2、3、4波段的伪彩色复合图像ETM234进行非监督分类，并通过计算分离率（TD值）对分类效果进行评价；同时将分类结果图与江宁县江滩钉螺孳生地矢量图重叠，提取江宁县江滩钉螺孳生地的植被特征，并分析钉螺在各植被中的分布。结果：通过对ETM234进行非监督分类，可将江宁县江潍地表特征分为包括水、沙滩、裸露地表等非植被区域在内的10种地表类型；对分类分离率评价显示，虽然总体分离率较好（TD＝1860），但存在类间的混杂错分。为了进一步提高分类效率，以值被指数为参考滤掉ETM234图像中无植被的像素然后再进行分类，消除了非相邻类间的混杂错分。提取江宁县江滩钉螺孳生地植被类型发现，钉螺孳生地的植被主要为非监督分类的第3（3）、5（C5）、6（C6）三类，且钉螺密度为C3＜C5＜C6，实地考察显示该三类分别为植被稀疏的滩面、杂草丛及芦苇滩。结论：通过对遥感图像适当分析能将江滩钉螺孳生地的植被种类有效地鉴别出来，有利于对钉螺孳生地的监测，为钉螺的有效防制及血吸虫病的预防提供依据。

Unsupervised classification of remote sensing image analysis of jiangning county river beach oncomelania breeding grounds of vegetation characteristics

Objective To explore the vegetation landscapes in marshland snail habitats using satellite image. Methods The false color composition image from band 2, 3 and 4 of LANDSAT ETM+ images was classified in the ERDAS IMAGINE 8.5 to analyze the vegetation types in the marshland of Jiangning county, using the unsupervised classification. The efficiency of classification was evaluated by the transformed divergence. The overlaid layers of the classified vegetation image and the vector layer of snail habitats distribution were used to analyze the relationship between the snail distribution and the landscape types. Results The land cover of marshland in LANDSAT ETM234 image in Jiangning county could be classified into 10 types, including water, bare soil, sandy and other landscapes while the transformed divergence analysis showed that there were misclassified pixes between some types especially for the non continuous types. The study indicated that through adding the NDVI image in the process of classification efficiency of classification and eliminate misclassification in the non continuous type could be improved. Analysis on the overlaid layer of the vector of snail distribution and the classified image proved that the vegetation covers in marshland snail habitats in Jiangning mainly belonged to type 3, 5 and 6, that responded to the beach with sparse vegetation,exuberant weed and bulrush respectively. The density of snails in the bulrush was higher than that in other 2 landscapes. Conclusion The vegetation type in the marshland snail habitats could be distinguished from the satellite image, which was helpful for the surveillance of snail habitat in marshland and for the prevention of schistosomiasis.