基于GF-1 PMS影像的土地利用/覆盖分类研究

Research on Land Use/Cover Classification Using GF-1 PMS Satellite Images

土地利用/覆盖变化是当前全球环境变化的重要内容之一,而土地利用/覆盖分类是其基础工作之一。以萍乡市区为研究区,利用国产GF-1 PMS影像为基础数据,采用最小距离、最大似然法、平行六面体和支持向量机四种监督分类法进行了土地利用/覆盖分类试验研究。结果表明:最大似然法和支持向量机分类算法具有更好的分类精度,总体精度分别为93.3和96.03,Kappa系数分别为0.917 2和0.948 7,而最小距离法和平行六面体的精度则差很多。由于GF-1 PMS的多光谱波段和全色波段的空间分辨率分别为8 m和2 m,因而支持向量机分类结果可以满足很多水文、生态等模型的需要。

Land use classification is an important basic work for global environment change study.This article presents a case study focused on land use classification in Pingxiang City,Jiangxi Province. Four different supervised classification methods were tested in this study,they are the minimum distance algorithm,the parallelepiped algorithm,the maximum likelihood algorithm and the support vector machine（ SVM） classifier. The result shows that the SVM is the best algorithm and the maximum likelihood algorithm followed it,the other two methods have more uncertainty. The overall classification accuracy and Kappa coefficient of SVM classifier is 96. 03 and 94. 87 respectively. Considering the advantages of high spatial resolution and short revisit period of GF-1 satellite,GF-1 PMS images can be important data resource for global environment change study.