地球信息科学学报
  欢迎光临地球信息科学学报期刊网
   
设为首页
加入收藏
 首页  |  期刊介绍  |  编 委 会  |  投稿须知  |  期刊订阅  |  文档下载  |  广告合作  |  联系我们  |  留言板  |  进入旧版
地球信息科学学报 2010, Vol. 12 Issue (3) :444-450    DOI:
本期要文(可全文下载)    
小波变换与FCM聚类的QuickBird影像特征空间识别分析
寇程, 柯长青
南京大学地理信息科学系, 南京 210093
Feature Space Recognition Analysis of QuickBird Imagery Based on Wavelet Transform and FCM Clustering
KOU Cheng, KE Changqing
Department of Geographical Information Science,Nanjing University,Nanjing 210093,China

摘要
参考文献
相关文章
Download: PDF (1378KB)   HTML 1KB   Export: BibTeX or EndNote (RIS)      Supporting Info
摘要 遥感影像分类是应用遥感影像进行地学分析等其他应用的重要准备工作,它的精度将直接影响到后续的分析工作。纹理特征对于提高高分辨率遥感影像的分类精度具有重要意义,小波变换的"时-频"分析方法在遥感影像纹理分析方面有着独特的优势。本文采用QuickBird影像,利用小波分解得到影像的纹理特征,结合光谱响应值组成特征空间,再利用模糊C均值聚类方法对影像进行分类。实验结果显示:加入了纹理特征的影像分类精度有所提高,同时,同一类地物的内部均一性有所改善。小波分析对于细微纹理特征的提取效果比粗纹理要好。
Service
把本文推荐给朋友
加入我的书架
加入引用管理器
Email Alert
RSS
作者相关文章
寇程
柯长青
关键词小波变换   FCM聚类   纹理特征   QuickBird影像     
Abstract: The remotely sensed imagery classification is an important preparation work for geo-science analysis with remotely sensed imagery.The accuracy of classification will influence the quality of the result of geo-science analysis.We can hardly achieve satisfactory classification result using traditional classification method in high-resolution remotely sensed imagery classification.The research shows that the texture is a very significant feature for improving the accuracy of classification of high-resolution remotely sensed imagery.The time-frequency method of wavelet transform has the spatial advantage in texture analysis of remotely sensed imagery.In this paper,the texture feature of remotely sensed imagery is extracted from QuickBird imagery via wavelet decomposing.The feature space is composed of texture feature and spectrum response value of every band of the image.Then the image is classified by fuzzy C mean-value clustering(FCM clustering) approach.The experimental result indicates that the accuracy of classification can be increased when the texture feature is added into feature space.Meanwhile,the inner-homogeneity of a single category is improved.We can also see that the wavelet analysis can extract subtle texture better than coarse texture.In this paper,it is from the gradient image that the texture feature is extracted,so there are some misclassified pixels at the edge of the image,like the sides of roads.Some tiny objects can be amplified.The change of parameters used in this method can impact on the result of classification,such as the window size of the wavelet mode variance,the window size of mean filter,the coefficients of spectral responses,and so on.How to find out the influence mechanism of these parameters and how to choose proper parameters are what need to be resolved in next researches.
Keywordswavelet transform   FCM clustering   texture feature   QuickBird imagery     
Received 2009-06-12;
Fund:

国家自然科学基金项目(40971044)

About author: 寇程(1987-),男,陕西人,硕士研究生。E-mail:koucheng@126.com;柯长青(1969-),男,陕西人,教授。研究方向为遥感、GIS、GPS应用。E-mail:kecq@nju.edu.cn
引用本文:   
寇程, 柯长青.小波变换与FCM聚类的QuickBird影像特征空间识别分析[J]  地球信息科学学报, 2010,V12(3): 444-450
KOU Cheng, KE Changqing.Feature Space Recognition Analysis of QuickBird Imagery Based on Wavelet Transform and FCM Clustering[J]  , 2010,V12(3): 444-450
Copyright 2010 by 地球信息科学学报