›› 2013, Vol. 42 ›› Issue (3): 0-0.

• 学术论文 •    下一篇



  1. 北京师范大学
  • 收稿日期:2012-02-22 修回日期:2012-11-30 出版日期:2013-06-20 发布日期:2014-01-23
  • 通讯作者: 慎利 E-mail:leeshenli@outlook.com
  • 基金资助:

River Extraction from the High Resolution Remote Sensing Image Based on Spatially Correlated Pixels Template and Adboost Algorithm

  • Received:2012-02-22 Revised:2012-11-30 Online:2013-06-20 Published:2014-01-23

摘要: 高分辨率遥感影像分析中,如何充分挖掘像素间的空间关系来保证分析结果的空间连续性,成为提高影像解译精度的关键。提出了一种度量像素间空间相关关系特征的邻域模式,即空间像素模板,并结合Adaboost集成学习算法来实现IKONOS高分辨率影像上河流的精确提取。首先,基于过滤式特征选择方法自动生成像素模板的具体形式,继而构建多维特征向量,然后利用Adaboost算法实现多特征的加权集成利用,从而最终精确地提取河流。相关实验结果表明,利用文中所提方法的河流提取结果较SVM方法、最大似然方法对应结果的精度更高,且面向对象特性更加明显。

关键词: 空间像素模板, 高分辨率遥感影像, Adboost算法, 河流提取

Abstract: For the analysis of high resolution remote sensing images, in addition to pixel-based spectral data, spatial and structural information should also be consider for a more intuitive and accurate interpretation. This paper presents a kind of neighbor patterns, referred to the spatially correlated pixels template, to incorporate the spatial relationship among pixels. And in conjunction with Adboost ensemble learning algorithm, accurate river extraction from an IKONOS high resolution remote sensing image is obtained. Firstly, a particular form of spatially correlated pixels templates is generated by using the filter feature selection approach. Secondly, the multi-dimensional feature vectors are constructed according to the given template. Then, Adboost algorithm is used to make full use of available features. Finally, the accurate river extraction is achieved by ensemble learning. Experimental results show the proposed methodology compares favorably with the SVM and maximum likelihood methods, particularly in the aspect of overall accuracy and object-oriented property.

Key words: Spatially Correlated Pixels Template, High Resolution Remote Sensing Image, Adboost Algorithm, River Extraction