地理研究

地理研究 ›› 2006, Vol. 25 ›› Issue (4): 579-586.doi: 10.11821/yj2006040003

• 论文 • 上一篇    下一篇

东亚飞蝗生境的遥感分类——以河北省黄骅地区为例

李开丽1,2, 倪绍祥1   

  1. 1. 南京师范大学地理科学学院,南京210097;
    2. 中国科学院南京土壤研究所,南京210097
  • 收稿日期:2005-12-08 修回日期:2006-05-25 出版日期:2006-08-15 发布日期:2006-08-15
  • 作者简介:李开丽(1977-),女,山东临沂人,博士研究生。主要从事遥感与GIS应用研究。 E-mail:lklcelery@163.com
  • 基金资助:

    国家自然科学基金资助项目(遥感与GIS支持的东亚飞蝗发生机理与预测模型研究,40371081)

Breeding area classification for oriental migratory locust assisted by remote sensing:a case study of the Huanghua Region in Hebei Province

LI Kai-li1,2, NI Shao-xiang1   

  1. 1. College of Geographical Science,Nanjing Normal University,Nanjing 210097,China;
    2. Insititute of Soil Science,CAS,Nanjing 210008,China
  • Received:2005-12-08 Revised:2006-05-25 Online:2006-08-15 Published:2006-08-15

PDF (PC)

赞

摘要：

东亚飞蝗生境的分类研究是东亚飞蝗监测和防治的一项重要基础工作。本文以河北省黄骅地区为研究区,基于两个时相的TM图像,采用三种遥感波段组合方案,以及最大似然分类和基于知识的分层分类两种分类方法,进行了东亚飞蝗生境的分类研究。结果表明,三种组合方案的分类总精度相差不大,其中加入图像纹理信息的最大似然分类法的分类总精度最高。但是,基于知识的分层分类法的分类精度在各单项生境类型之间相差较小,从而显示出该方法在应用上仍有一定的优越性。

关键词: 遥感, 图像分类, 生境, 东亚飞蝗

Abstract:

The classification of breeding area for oriental migratory locust（Locust migratoria manilensis Meyen） is one of the most important tasks in terms of the monitoring and controlling of the damages induced by the locusts.In this study,the Huanghua region along the Bohai Bay in Hebei Province was selected as the study area and the locust breeding areas were classified based on the Landsat-5 TM images dated on August 14,2003（TMⅠ） and May 28,2004（TMⅡ） respectively.Three different schemes of image band combination and two kinds of classifiers were used in the breeding area classification,i.e.the maximum likelihood classifier and the knowledge-based layered classifier.In more detail,they are 1) the combination of bands 3,4 and 5 of TMⅠplus bands 3,4 and 5 of TMⅡwith the maximum likelihood classifier;2)the combination of bands 3,4 and 5 of TMⅠplus bands 3,4 and 5 of TMⅡand the homogeneity index derived from the image of NDVITMⅠas a band which contains the spatial texture information of the images,with also the maximum likelihood classifier;and 3)the combination of bands 3,4 and 5 of TMⅠplus bands 3,4 and 5 of TMⅡand the NDVITMⅠas a band,with the knowledge-based layered classifier.The results show that,firstly,there is no obvious difference among these different combination schemes in terms of the overall accuracy of the locust breeding area.Relatively speaking,the overall accuracy of the second combination scheme（89.319） is somewhat higher than those of the other two combination schemes,which indicates that it is beneficial to accuracy improvement of locust breeding area classification if adding the spatial texture information of the images into the classification.Secondly,although the overall accuracy of locust breeding area classification with the third combination scheme is somewhat lower than those of the other two combination schemes,its variation range of locust breeding area classification accuracy among all individual locust breeding area types is relatively small,which means that the knowledge-based layered classifier still has a certain of advantages in the locust breeding area classification.

Key words: remote sensing, image classification, breeding area, oriental migratory locust