测绘学报

• 学术论文 • 上一篇    下一篇

微粒群优化方法的遥感影像变化检测研究

戴芹,刘建波,刘士彬   

  1. 中国科学院对地观测与数字地球科学中心
  • 收稿日期:2011-06-27 修回日期:2012-01-18 出版日期:2012-12-25 发布日期:2013-04-17
  • 通讯作者: 戴芹

Remote Sensing Image Change Detection Using Particle Swarm Optimization Algorithm

  • Received:2011-06-27 Revised:2012-01-18 Online:2012-12-25 Published:2013-04-17

摘要:

针对目前多时相光谱直接变化检测方法存在训练样本分布限制和样本特征组合单一的缺陷问题,本文将微粒群优化方法引入遥感信息变化检测领域,构建了基于微粒群优化方法的遥感信息变化检测方法,在变化监测的过程中,通过变化规则的自动搜索和建立,实现了遥感影像变化信息的一次性直接提取。在方法验证过程中,选择北京为实验区,成功实现了应用微粒群优化方法对实验区2000年至2006年、2006年至2009年两个时间段的遥感影像进行了土地覆盖类型的变化信息检测,并将应用微粒群优化方法与决策树(C4.5和PART)、最大似然等方法的变化检测结果进行了对比分析。结果表明,微粒群优化方法能够自动搜索变化规则,得到的变化规则比决策树方法更简单,并能够获得更高的检测精度。

关键词: 微粒群优化, 多时相遥感影像, 变化检测

Abstract:

Aiming at the restriction of training samples distribution and limitation of feature combinations caused by traditional methods in the current multi-temporal remote sensing image direct change detection application, this paper introduces the particle swarm optimization(PSO)method to the field of remote sensing change detection, and propose a new change detection method based on PSO algorithm. In the processing of change detection, it automatically searches the change rules, so it can directly achieve the change information at one time. Selecting Beijing area as experimental area, this paper demonstrates the land cover change detection information extraction in Beijing area from 2000 to 2006, 2006 to 2009 using the new method. The PSO method is also compared with C4.5, PART, Maximum Likelihood methods, the results show that the PSO algorithm can search change rules automatically, and can achieve simpler rule than C4.5 and PART, can achieve high precision than the other three methods.

Key words: Particle Swarm optimization algorithm, multi-temporal remote sensing image, change detection