基于面向对象分类方法的Rapideye桉树提取

摘要

本研究利用面向对象分类方法和Rapideye高分辨率遥感影像对广东省肇庆市部分地区进行自动遥感分类进而提取桉树空间分布信息，并将提取结果与传统的监督分类方法(最小距离法)提取结果进行对比分析。本研究主要对桉树等5种林地类型以及草地、其它用地等共7种地物类型进行分类提取。通过对比面向对象方法和最小距离法分类结果，发现基于面向对象分类方法提取桉树林地结果较好，分类精度可达82.08%，Kappa系数可达0.74，分类效果显著优于传统的监督分类方法，面向对象方法比较适合于桉树信息的分类提取。%In this study, the object-oriented classification method and Rapideye high resolution remote sensing images were used to examine the spatial distribution ofEucalyptus and compare to the traditional supervised classification method (minimum distance). This study involved classification of 7 land cover types, which included 5 classes of forest (includingEucalyptus plantations), grass and other land uses. By comparing the results of objected-oriented method with minimum distance method, we found that objection-oriented classification showed obviously better performance on extractingEucalyptus plantation cover measures and classification of forest type than the minimum distance method, its accuracy of general classification can reach 82.1% and its Kappa coefficient can reach 0.74.