测绘学报

• 学术论文 •    

基于小波变换的多时相SAR图像变化检测技术

黄世奇,刘代志,胡明星,王仕成   

  1. 第二炮兵工程学院
  • 收稿日期:2009-01-23 修回日期:2009-07-15 出版日期:2010-04-22 发布日期:2010-04-22
  • 通讯作者: 黄世奇

Unsupervised Two-Threshold Change Detection Algorithm Based on Wavelet Transform in Multi-temporal SAR Images

  • Received:2009-01-23 Revised:2009-07-15 Online:2010-04-22 Published:2010-04-22

摘要: 多时相SAR图像变化检测的难点是SAR复杂的相干成像机理及其带来的斑点噪声。小波多尺度分解能对SAR图像斑点噪声的抑制和几何细节的保存起到权衡的作用,还能通过不同的尺度来描述SAR图像的变化。本文提出了无人监督的双阈值小波变换(UTWT)SAR图像变化检测算法。采用期望最大(EM)算法产生双阈值,可以区分像素发生变化的类型(变化区域增强类和变化区域减弱类)或变化等级。用SAR图像数据进行了实验,实验结果表明该方法有效。

Abstract: The difficulty of multi-temporal synthetic aperture radar (SAR) image change detection is that SAR coherence imaging mechanism is very complex and it brings speckle noise. Wavelet multi-scale decomposition can reach the balance effect for the reduction of SAR image speckle noise and the protection of geometry detail, furthermore, it can describe SAR image changes through different scales. This paper proposes the unsupervised two-threshold wavelet transform (UTWT) SAR image change detection algorithm. Using the expectation maximization (EM) algorithm produces two threshold values, which can distinguish change classes (the enhanced change region and the weakened change region) or change grades. Finally, some SAR images perform experiments and the experimental results verify that the method is effective.