测绘通报 ›› 2017, Vol. 0 ›› Issue (1): 65-68,73.doi: 10.13474/j.cnki.11-2246.2017.0014

• 学术研究 • 上一篇    下一篇

基于布谷鸟算法的遥感影像智能分类

沈泉飞1, 曹敏2, 史照良3, 许如琪2   

  1. 1. 江苏省基础地理信息中心, 江苏 南京 210013;
    2. 南京师范大学虚拟地理环境教育部重点实验室, 江苏 南京 210023;
    3. 江苏省测绘地理信息局, 江苏 南京 210013
  • 收稿日期:2015-10-27 修回日期:2016-06-28 出版日期:2017-01-25 发布日期:2017-02-06
  • 通讯作者: 曹敏
  • 作者简介:沈泉飞(1980-),男,硕士,工程师,研究方向为遥感影像处理。E-mail:sqf1980@126.com
  • 基金资助:
    国家自然科学基金(41101349);江苏省高校自然科学基础研究项目(13KJB420003);江苏高校优势学科建设工程资助项目(PAPD)

Intelligent Remote Sensing Classification Based on Cuckoo Search Algorithm

SHEN Quanfei1, CAO Min2, SHI Zhaoliang3, XU Ruqi2   

  1. 1. Provincial Fundamental Geomatics Centre of Jiangsu, Nanjing 210013, China;
    2. College of Geographic Science, Nanjing Normal University, Nanjing 210023, China;
    3. Jiangsu Provincial Bureau of Surveying, Mapping and Geoinformation, Nanjing 210013, China
  • Received:2015-10-27 Revised:2016-06-28 Online:2017-01-25 Published:2017-02-06

摘要: 提出了一种新的基于布谷鸟算法的智能式遥感分类方法。采用布谷鸟智能优化算法,自动搜索遥感影像各波段的最优阈值分割点,并定义各波段最优阈值分割点和影像分类目标类别的连线为布谷鸟的最佳解,构造以If-Then形式表达的遥感分类规则。将所提的基于布谷鸟算法的影像分类方法应用于ALOS影像分类中,并与蜂群智能遥感分类方法和See5.0决策树方法进行了对比分析。结果表明,布谷鸟智能遥感分类的总体精度和Kappa系数均比蜂群智能遥感分类和See5.0决策树方法更高,该智能遥感分类方法具有更好的分类效果。

关键词: 布谷鸟算法, 仿生智能计算, 遥感影像, 分类

Abstract: A new, intelligent approach to classify remote-sensing images based on Cuckoo search algorithm is presented. Cuckoo search algorithm, a new bio-inspired intelligence algorithm, is widely used to solve optimization problems. Cuckoo search algorithm to search for the optimal upper and lower threshold values on each band of remote-sensing image is applied. The classification rules are constructed by the links between the optimal split values and classification type in the explicit formation of ‘if-then’, and each link corresponds to the optimal solution of one Cuckoo, nest or egg. By taking an example of ALOS image in the north shore of the Yangtze River estuary, the proposed classification method based on CS algorithm is implemented and tested against See5.0 decision-tree method. The overall classification accuracy and Kappa coefficient of CS-based method are higher than the See5.0 decision-tree one. The results demonstrate that the practicability of applying CS algorithm to classify the remote-sensing images.

Key words: cuckoo search algorithm (CS), swarm intelligence caculation, remote sensing image, classification

中图分类号: