测绘学报 ›› 2014, Vol. 43 ›› Issue (8): 842-847.doi: j.cnki.11-2089.2014.0131

• 学术论文 • 上一篇    下一篇

纹理特征与视觉注意相结合的建筑区提取

沈小乐1,邵振峰2,田英洁1   

  1. 1. 武汉大学
    2. 武汉大学测绘遥感信息工程国家重点实验室
  • 收稿日期:2013-12-02 修回日期:2014-04-11 出版日期:2014-08-20 发布日期:2014-08-27
  • 通讯作者: 沈小乐 E-mail:shen_xiaoxiaole@163.com

Built-Up Areas Extraction of High-Resolution Remote Sensing Images by Texture Driven Visual Attention Mechanism

  • Received:2013-12-02 Revised:2014-04-11 Online:2014-08-20 Published:2014-08-27

摘要:

本文提出一种有效提取高分辨率遥感影像建筑区的算法。作者通过模拟人类视觉系统感知环境的过程,引入视觉注意机制,提出了基于自信息最大化视觉注意机制的建筑区显著指数。其中根据建筑区在高分辨率遥感影像中的特点,基于时/频域的纹理分析,提出了一种针对建筑区的纹理特征描述方法,实现了建筑区的高效提取。通过对8组高分辨率遥感影像进行实验,验证了本文算法的有效性。

关键词: 高分辨率遥感影像, NSCT, 纹理特征, 视觉显著性, 建筑区提取

Abstract:

An effective built-up areas extraction algorithm of high-resolution remote sensing images is proposed in this paper. Based on a signal processing method, we present a texture description method for the characteristics of built-up areas in high-resolution remote sensing images. Through an information theory based visual attention mechanism, a saliency map is calculated to extract built-up areas. Experiments prove the effectiveness of the proposed algorithm.

Key words: high-resolution remote sensing images, NSCT, textural features, visual saliency, built-up areas extraction

中图分类号: