首页    编委会    编辑部    历任主编    获奖信息    征稿简则    友情链接    English    旧版主页
  在线期刊
       最新录用
       当期目次
       过刊浏览

  下载专区
       版权协议
       征稿范文
       来稿登记表
       录用证明

  联系我们

投稿e-mail: jbnuns_sub@bnu.edu.cn

主编e-mail: jbnuns_eic@bnu.edu.cn

办公地点:北京师范大学 科技楼A区216,218,220室

电话:(010)58807851,58802050,58807872,58807713

通信处:北京市海淀区新街口外大街19号北京师范大学理科学报编辑部

邮政编码:100875

办公室e-mail: jbnuns@bnu.edu.cn

本网站暂不提供全文下载,如需全文下载请点击

中国知网

 

标题

随机森林在高光谱遥感数据中降维与分类的应用

作者

杨珺雯 张锦水 朱秀芳 谢登峰 袁周米琪

机构

北京师范大学地表过程与资源生态国家重点实验室;北京师范大学资源学院

摘要

高光谱数据的特征数目庞大,而且波段之间存在冗余信息,对高光谱数据进行分类的成本较高,因此需要提取合适的特征达到提高效率的目的。随机森林作为一种热门算法,广泛应用于各种分类、特征选择等问题中,均取得了良好的效果。本文选择北京小汤山农业试验区的OMIS高光谱影像作为研究数据,利用随机森林算法计算每个特征的重要性指标并对其排序,针对面向精度和效率的特征选择策略,使用RF-RFE波段选择方法去除价值低的特征分别得到最佳波段组合,实现高光谱数据降维,进行随机森林、支持向量机分类。实验结果表明随机森林分类精度为72.82%,SVM分类精度为65.21%,随机森林分类器优于SVM,证明了其为良好的高光谱数据分类器。

关键词

OMIS,高光谱,随机森林,RF-RFE,降维,波段选择,分类

引用

杨珺雯, 张锦水, 朱秀芳, 谢登峰, 袁周米琪. 随机森林在高光谱遥感数据中降维与分类的应用[J]. 北京师范大学学报(自然科学版),2015,51(Sup.1):82-88.

基金

国家自然科学基金资助项目(41301444);国家高分辨率对地观测系统重大基金资助项目(民用部分);北京高等学校“青年英才计划”基金资助 项目;北京市自然科学基金资助项目(8144052)

分类号

TP751

DOI

10.16360/j.cnki.jbnuns.2015.s1.013

Title

Random forest applied for dimension reduction and classification in hyperspectral data

Author

YANG Junwen, ZHANG Jinshui , ZHU Xiufang, XIE Dengfeng Xie, YUAN Zhoumiqi

Affiliations

College of Resources Science & Technology, Beijing Normal University; State Key Laboratory of Earth Surface Processes and Resource Ecology College of Resources Science and Technology, Beijing Normal University

Abstract

Hyperspectral image contains a huge amount of data, there is great redundancy information in hyperspectral data and band selection can remove it effectively and reduce computational cost accordingly. As a type of integrated learning method, Random Forest algorithm has been applied to classification and feature extraction, and great results have been made. OMIS hyperspectral image of experimental plot in Xiaotangshan, Beijing was used in the present work. Random forest algorithm was used to calculate the value of each feature index. For feature selection strategy oriented for efficiency and accuracy, the feature with low value was removed to obtain optimum combination of bands. Accuracy in the Random forest classifier was found to be up to 72.82%, the SVM classification accuracy was 65.21%. Therefore Random Forest algorithm could adapt to hyperspectral image data and has better precision than SVM.

Key words

OMIS; hyperspectral; Random Forest; RF-RFE; dimensionality reduction; classification

cite

YANG Junwen, ZHANG Jinshui , ZHU Xiufang, XIE Dengfeng Xie, YUAN Zhoumiqi.Random forest applied for dimension reduction and classification in hyperspectral data [J]. Journal of Beijing Normal University(Natural Science),2015,51(Sup.1):82-88.

DOI

10.16360/j.cnki.jbnuns.2015.s1.013

Copyright © 2014 Journal of Beijing Normal University (Natural Science)
Designed by Mr. Sun Chumin. Email: cmsun@mail.bnu.edu.cn