基于高分辨率遥感影像的河套灌区耕地信息提取研究

利用libsvm软件包的交叉验证方法选择了支持向量机的最优参数,利用ENVI5.1软件将支持向量机作为分类器,对覆盖整个河套灌域的2014年5-9月的高分一号和资源三号卫星遥感影像进行监督分类,提取出灌区耕地信息,经过人工修改,最终得到河套灌区耕地面积为634804.81ha,几何精度为91.23%,属性精度为93%。

The optimal parameters of support vector machine are selected using the cross-verification method of LIBSVM software package, and the support vector machine is used as the classifier to supervise and classify the satellite remote sensing images of Gaofen No.1 and Resource No.3, which cover the whole Hetao irrigation area from May to September 2014. The cultivated land information is extracted. After manual modification, the cultivated land area of Hetao irrigation area is 634804.81ha, the geometric accuracy is 91.23% and the attribute accuracy is 93%.