基于感兴趣区域在区域层面上决策的SAR图像变化检测的方法研究

在现有的变化检测方法中，针对差分图像的方法得到了广泛的应用，但现有该框架下的技术都在在像素层面上决策产生的变化检测结果，这易使结果中存在诸如有噪声般散落的杂点、连通区域内有孔洞及边缘不平滑等缺陷。为此，本文给出一种在区域层面上决策生成变化检测结果的技术，其核心为抽取并处理感兴趣区域，而关键在于获取合适的抽取感兴趣区域的标签和如何在区域层面上生成变化检测结果。为使抽取的感兴趣区域包含几乎所有的变化类信息，我们用平稳小波变换和模糊C-均值(Fuzzy C-Means, FCM)算法分两步获取抽取感兴趣区域的标签；为在区域层面上生成变化检测结果，我们依据标签搜索感兴趣区域内所有的连通区域，并把每个连通区域看作为一个数据点，再由阈值技术处理这些数据点生成最终的变化检测结果。对真实SAR图像数据集的变化检测结果表明，其主观效果和客观性能都优于其他相关技术的。

A Region-of-interest Level Method for Change Detection in SAR Imagery

A novel approach which works at the region level is proposed based on extracting and handling region of interest(ROI).It contains two key points.One is to obtain a proper label for extracting ROI by combining stationary wavelet transform(SWT),fuzzy C-means(FCM) clustering algorithm with inter-scale fusion strategy.The other is to generate the final change detection map at the region level through the act of searching all connected regions and then looking at each one as a unit to handle.In virtue of both points,this method has both quantitative and qualitative performances superior over other related methods.Results tested on the real SAR datasets have also confirmed this.