为研究不同面向对象分类方法对GF-2影像桉树信息提取的可靠性,以广西平朗乡为研究区,首先对GF-2影像做多尺度分割处理,然后采用贝叶斯(Bayes)、决策树(DST)、K最近邻法(KNN)、支持向量机(SVM)、随机森林(RDT)等5种分类器进行分类提取,最后对提取结果进行混淆矩阵精度分析。结果表明:在这5种分类方法中,支持向量机方法的精度最高,总体精度达86.4%,Kappa系数为0.73,贝叶斯方法的精度最差。可见GF-2影像可以作为桉树信息遥感监测的数据源之一,且支持向量机分类方法是桉树提取的较优选择。

In order to explore GF-2 satellite performance in eucalyptus information extraction,this research took place in Pinglang County.After multi-scale segmentation,the research used object-oriented classification methods like Bayes(Bayes),Decision Tree(DST),K nearest neighbor(KNN),Support Vector Machine(SVM)and Random trees(RDT)to extract the Eucalyptus information.Then it evaluated the classification accuracy by Confusion Matrix.The results showed that the Support Vector Machine was the best classification method with the overall accuracy at 86.4%,and the Kappa coefficient 0.73.The Bayes classification was worst.Thus,GF-2 satellite data can provide a potential data source for remote sensing monitoring of eucalyptus information,and the Support Vector Machine would be a potential method.