首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于支持向量机的MERIS土地覆盖制图及其空间一致性分析
引用本文:李明诗,Chandra Giri,朱智良,吕恒,潘洁,温卫松,徐达,刘安兴.基于支持向量机的MERIS土地覆盖制图及其空间一致性分析[J].南京林业大学学报(自然科学版),2008,32(6).
作者姓名:李明诗  Chandra Giri  朱智良  吕恒  潘洁  温卫松  徐达  刘安兴
作者单位:1. 南京林业大学森林资源与环境学院,江苏,南京,210037
2. 美国地质调查局EROS数据中心,苏福尔斯,SD,57198
3. 美国林务局,阿灵顿,VA,22209
4. 南京师范大学,江苏省地理信息科学重点实验室,江苏,南京,210097
5. 浙江省森林资源监测中心,浙江,杭州,310020
摘    要:遵循美国国家土地覆盖数据库2001分类主题及系统(30 m空间分辨率),研究中等分辨率成像光谱辐射仪MERIS(300 m)土地覆盖产品的发展及评价。4种监督分类器包括马氏距离、最大似然、决策树以及支持向量机被用来发展区域土地覆盖信息。结果表明:(1)支持向量机在土地特征刻画过程中分类性能最优;(2)由支持向量机导出的MERIS土地覆盖产品尽管其识别地面细节的能力不及NLCD2001,但其主要地物类型在空间分布上与NLCD2001比较接近。分析还进一步揭示MERIS数据可成功地区划水体、常绿森林、裸地及栽培作物等地物类型,而对于落叶林及灌木林的刻画则性能相对较差。在MERIS土地覆盖产品中观察到从灌木林向裸地、灌木林向常绿森林及灌木林向草地的误分现象。然而,MERIS土地覆盖产品的生产较NLCD2001要节省人力及成本,中等尺度的MERIS土地覆盖产品对于某些科学应用将具有独特的价值。MERIS土地覆盖产品的发展应该充分应用多种辅助信息以及区域调制的分类策略,以期获得更加可靠的分类结果。

关 键 词:决策树  支持向量机  空间一致性分析

Use of support vector machines algorithm to map MERIS land cover and its spatial agreement analysis
LI Ming-shi,Chandra Giri,ZHU Zhi-liang,L Heng,PAN Jie,WEN Wei-song,XU Da,LIU An-xing.Use of support vector machines algorithm to map MERIS land cover and its spatial agreement analysis[J].Journal of Nanjing Forestry University(Natural Sciences ),2008,32(6).
Authors:LI Ming-shi  Chandra Giri  ZHU Zhi-liang  L Heng  PAN Jie  WEN Wei-song  XU Da  LIU An-xing
Institution:LI Ming-shi,Chandra Giri,ZHU Zhi-liang,L(U) Heng,PAN Jie,WEN Wei-song,XU Da,LIU An-xing
Abstract:This study focused on the development and assessment of the Medium Resolution Imaging Spectrometer (MERIS) land cover product.Four supervised classifiers including the Mahalanobis distance,maximum likelihood,decision trees and support vector machines (SVM) were applied to develop land cover information following the National Land Cover Database(NLCD) 2001 classification scheme.Results showed that SVM algorithm performed most optimally.The derived MERIS land cover was spatially close to NLCD 2001,although its capability for identifying ground details was less powerful than NLCD 2001.Furthermore,MERIS data were successful at delineating water,evergreen forest,barren land and cultivated crops,and less successful at characterizing deciduous forest and shrub/scrub.Misclassification of shrub/scrub to barren land,evergreen forest,and grassland were observed in MERIS land cover.However,production of MERIS land cover is much less labor-intensive and cost-effective than that of NLCD 2001,so the moderate resolution MERIS land cover may have value for specific applications.Future production of MERIS land cover should adequately use diverse ancillary information and a regionally tuned classification strategy to achieve more reliable results.
Keywords:MEIRS  MERIS  Decision trees  Support vector machines  Spatial agreement analysis
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号