摘要
以提取乌伦古湖湿地信息为目的,采用该地区2010年Landsat5/TM影像,使用支持向量机SVM的多项式函数对湿地信息进行分类。使用精度评估法(混淆矩阵)对分类结果进行精度评价并与传统的最大似然法(MLC)、非监督分类(ISODATA)法分类结果进行对比。结果表明,SVM分类法不仅能够很好地提取河流信息并且能够区分湖泊湿地与坑塘湿地,分类总精度达到94.000%,Kappa系数为0.932,明显高于MLC、ISODATA两种方法,同时各类别的用户精度和制图精度都在不同程度上高于传统分类方法。SVM是一种有效的提取湿地的方法 ,非常适用于干旱区湿地信息提取与监测。
In order to extract the information of Ulungur Lake Wetland using Landsat5 / TM image in 2010 and the polynomial function of support vector machine（SVM） to get the information classification of study area. Using the accuracy assessment method（the confusion matrix） to evaluate the accuracy of classification results,and compared the classification results with the traditional methods such as maximum likelihood method（MLC） and the unsupervised classification method（ISODATA）. The results showed that, the SVM classification method could not only extract the river information well,but also distinguish the lake wetland and pond wetland, and it＇s overall classification accuracy was 94.000% and the Kappa coefficient was up to0.932,both significantly higher than that of MLC and ISODATA. At the same time, various categories of users precision and product accuracy were in varying degrees higher than that of the traditional classification method. So,SVM was an effective method to extract wetland,and was a suitable method for the wetland information extraction and monitoring in arid area.
出处
《湖北农业科学》
2016年第16期4145-4149,共5页
Hubei Agricultural Sciences
关键词
SVM
遥感
干旱区
湿地
SVM
remote sensing
arid areas
wetlands
