计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (20): 23-25.DOI: 10.3778/j.issn.1002-8331.2008.20.008

• 博士论坛 • 上一篇    下一篇

基于AdaBoost的高分辨率遥感影像城市绿地提取算法

许 凯1,秦 昆1,杜 鹢2   

  1. 1.武汉大学 遥感信息工程学院,武汉 430079
    2.通信网络技术管理中心,北京 100840
  • 收稿日期:2008-03-12 修回日期:2008-04-22 出版日期:2008-07-11 发布日期:2008-07-11
  • 通讯作者: 许 凯

New algorithm of extraction of urban vegetation from high resolution image based on AdaBoost

XU Kai1,QIN Kun1,DU Yi2   

  1. 1.Information Engineering of Remote Sensing,Wuhan University,Wuhan 430079,China
    2.Communication Network Technology Management Center,Beijing 100840,China
  • Received:2008-03-12 Revised:2008-04-22 Online:2008-07-11 Published:2008-07-11
  • Contact: XU Kai

摘要: 针对高分辨率遥感影像中城市绿地的光谱、纹理及NDVI等特征,提出了一种基于多特征、多检测器组合的高分辨率遥感影像绿地提取算法。该算法分别以城市绿地的光谱信息、纹理及NDVI来构造提取绿地的弱检测器,并通过AdaBoost算法进行训练,将弱检测器加权组合构成提取绿地的强检测器。对武汉地区Quick Bird影像进行了绿地提取实验,实验结果表明该方法可以高效并准确地提取出绿地信息,优于传统的单一采用光谱信息或纹理信息的绿地提取算法。

关键词: AdaBoost, 绿地, Quick Bird, NDVI, 纹理

Abstract: This paper focuses on the extraction of urban vegetation from high resolution remote sensing image.A new algorithm of segmentation of urban vegetation from Quick Bird image based on AdaBoost is proposed.The new algorithm combines all weak detectors which are constructed by spectral images,NDVI and textures to form a strong vegetation detector.Finally,the experiments show that this method can extract vegetation in Quick Bird image exactly,and more effectively than traditional algorithm.

Key words: AdaBoost, vegetation, Quick Bird, NDVI, texture