基于随机森林的大姚县TM遥感影像分类研究  

Classification of TM Remote Sensing Image Based on Random Forests of Dayao County

作　　者：王栋[1] 岳彩荣[1] 田传召 范怀刚 王跃辉[1] 

机构地区：[1]西南林业大学林学院,云南昆明650224

出　　处：《林业调查规划》2014年第2期1-5,共5页Forest Inventory and Planning

基　　金：国家自然基金(31260156);亚太森林网络(APFNET/2011/PA004);西南林业大学科技创新基金(1339)

摘　　要：随机森林(Random Forest)是一种组合多棵决策树分类器的新的分类算法。以楚雄州大姚县为例,采用Landsat-TM数据,通过最大似然、支持向量机、随机森林3种分类器进行分类对比研究。结果表明,支持向量机和随机森林的分类精度明显优于最大似然法,两者分类精度相差不大;在分类时间上,最大似然法明显比随机森林和支持向量机快,支持向量机最慢。综合分析,随机森林算法表现更优,它在保证分类精度的前提下,也能保证一定的时间效率,更适宜实际生产应用。Random Forest is a new classification algorithm,which is combined with a classifier of multiple decision trees. By using Landsat-TM data of Dayao County,Chuxiong Province,and taking three classification methods including maximum likelihood,support vector machine and random forest,the superiority of random forest classifier are analyzed. The results show that the classification precision of support vector machine and random forest classification is obviously superior to the maximum likelihood,and they have little difference in the precision. At the classified time,maximum likelihood is significantly faster than random forests and support vector machine,and support vector machine is the slowest one. According to comprehensive analysis,random forest method is the best one,it does not only ensure the classified precision,but also guarantee efficiency; it is more suitable for actual production applications.

关 键 词：随机森林 遥感影像 分类精度 大姚县 

分 类 号：S771.8[农业科学—森林工程]