地震地质 ›› 2009, Vol. 31 ›› Issue (4): 607-616.DOI: 10.3969/j.issn.0253-4967.2009.04.004

• 活动火山监测与研究 • 上一篇    下一篇

基于遥感影像光谱与纹理分析的地物分类——以长白山天池火山地区为例

许建东1, 栾鹏2, 樊笑英3, 林旭东1   

  1. 1. 中国地震局地质研究所, 北京, 100029;
    2. 中国国士资源经济研究院, 北京, 101149;
    3. 中国国土资源部咨询研究中心, 北京, 100035
  • 收稿日期:2009-07-13 修回日期:2009-11-02 出版日期:2009-12-24 发布日期:2010-04-17
  • 作者简介:许建东,男,1962年生,毕业于美国纽约州立大学布法岁分校地质系,获得博士学位,研究员,从事火山地质、火山灾害研究,电话:010-62009155,E-mail:jdx321@sina.com.
  • 基金资助:
    “十一五”国家科技支撑计划项目(2006BAC01B040102);地震行业科研专项(200708027)共同资助

ANALYSIS OF SPECTRUM AND TEXTURE INFORMATION ON CHANGBAISHAN TIANCHI VOLCANO CALDERA AND ITS APPLICATION

XU Jian-dong1, LUAN Peng2, FAN Xiao-Ying3, LIN Xu-dong1   

  1. 1. Institute of Geology, China Earthquake Administration, Beijing 100029, China;
    2. Chinese Academy of Land and Resources Economics, Beijing 101149, China;
    3. The Consulting and Research Center, Ministry of Land and Resources, Beijing 100035,China
  • Received:2009-07-13 Revised:2009-11-02 Online:2009-12-24 Published:2010-04-17

摘要: 利用遥感影像的光谱分析和纹理分析技术,研究了长白山天池火山地区IKONOS影像的光谱特征以及纹理特征。对照野外实际调查结果,选取了8种训练样本类别,首先使用波谱角技术(SAM)和最大似然法进行分类,并通过混淆矩阵评价分类结果;然后,引入了ENVI的二阶概率矩阵法的纹理分析,利用差异性(Dissimilarity)纹理特征图对地物建立了纹理解译标志;最后,通过光谱和纹理结合的方法得出了最终的分类结果,使得分类精度得到了很大提高。以研究区广泛分布的浮岩为例,确定了浮岩的分布范围和相对厚度分布,分析了不同厚度浮岩的地形特征。

关键词: 遥感, 光谱分析, 纹理分析, 长白山天池

Abstract: The terrain near the Changbaishan Tianchi volcano caldera is complex and highly covered with vegetation,so it is difficult to determine the precise spatial distribution extent of the volcanic eruptive products only using the geological or visible remote sensing technology.In this paper,based on the analysis of spectrum and texture information from the IKONOS image of the study area,eight types of training samples are selected and the spectral angle technology(SAM)is applied to perform classification. These eight types of surface include water body,shadow,thick pumice,thin pumice,trachyte, soil,vegetation and forest.The primary results of such classification are not satisfactory through confusion matrix evaluation,which may be due to a serious loss of spectrum information in the synthesizing process.However,the texture information of the image is abundant.Therefore,we introduce the texture analysis of the ENVI probability matrix method which takes advantage of dissimilarity texture feature to establish the symbol of texture interpretation.Our final results indicate that the accuracy of classification may be greatly improved by using the combination of spectrum and texture analysis. Taking the widely distributed pumice as an example of classification,we obtain the information of both the extent and the relative thickness distribution of pumice near the Tianchi caldera region.In the future, more work should be concentrated on the field investigation to confirm the terrain features of eight types which are obtained from the satellite image interpretation before that remote sensing technology will truly become one of the useful tools for geological mapping and volcanic hazards evaluation.

Key words: remote sensing, spectrum analysis, texture analysis, Changbaishan Tianchi volcano

中图分类号: