欢迎访问《中国农学通报》，
中国农学通报

中国农学通报 ›› 2014, Vol. 30 ›› Issue (28): 61-66.doi: 10.11924/j.issn.1000-6850.2014-1168

• • 上一篇    下一篇

基于Landsat-8 影像森林植被信息计算机自动提取研究

张颖,王越男,陈利,刘秀英   

  1. 河南省广播电视大学,安化县林业局,中南林业科技大学林业遥感信息工程研究中心,河南科技大学农学院
  • 收稿日期:2014-04-21 修回日期:2014-04-21 接受日期:2014-08-18 出版日期:2014-10-15 发布日期:2014-10-15
  • 通讯作者: 陈利
  • 基金资助:
    河南省2013 年科技攻关项目“冬小麦主要生理指标的高光谱定量提取与应用研究”(132102110210)

Forest Vegetation Information Computer Automatic Extraction Base on Landsat-8

Ying ZHANG,, and   

  • Received:2014-04-21 Revised:2014-04-21 Accepted:2014-08-18 Online:2014-10-15 Published:2014-10-15

摘要： Landsat-8能够提供15 m全色波段和30 m分辨率的多光谱波段，Landsat-8上携带有OLI（operational land imager，陆地成像仪）和TIRS（thermal infrared sensor，热红外传感器）2个主要载荷，OLI陆地成像仪包括9个波段，TIRS包括2个热红处波段，全色波段Band8波段范围较窄，这种方式可以在全色图像上更好区分植被和无植被特征。本研究以攸县为例，采用Landsat-8遥感影像为数据源，进行缨帽变换及主成分分析处理，利用决策树分类模型进行提取。结果表明：Landsat-8遥感数据经过缨帽变换和主成分分析处理后，增强纹理信息，突出各地物的特征，把各地物在经过处理后的灰度值作为决策树分类模型的阈值，利用计算机自动提取，提取的总体精度为84.7%，攸县森林植被的面积为150911.7 hm2与以往的只利用波段的灰度值及植被指数等作为阈值相比，精度明显提高，方法也得到改善，得到了比较好的提取结果。

关键词: 青田, 青田

Abstract: Landsat-8 with two main load, OLI (operational land imager) and TIRS (thermal infrared sensor), can provide 15 m panchromatic band and 30 m resolution multispectral bands. OLI included nine bands, TIRS included two hot red bands, pan band 8 was narrower. This approach can better distinguish between vegetation and no vegetation characteristics on the full- color image. In this study, taking Youxian for example, after tasseled cap transformation and principal component analyzing and processing based on Landsat- 8 remote sensing image, vegetation was extracted by using decision tree classification model. The results showed that: the Landsat- 8 after tasseled cap transformation and principal component analyzing and processing could significantly enhance the texture information of the image, and highlight the land features. After tasseled cap transformation and processing of principal component analysis, put the gray values as a threshold of decision tree classification and used the computer automatic extraction, the overall accuracy was 84.7%, with an area of forest vegetation of 150911.7 hm2. Compared with the previous threshold band using only gray values and vegetation index, the accuracy was significantly increased and the method was also improved, which could led to better extraction results.