论文部分内容阅读
SRM(Statistical Region Merging)分割算法具有快速、稳定和抗噪强的优点,基于此,本文提出一种基于DSSRM(Dynamic Sorting Statistical Region Merging)级联分割的SAR图像变化检测方法。首先,针对SRM算法基于单特征静态排序导致的过分割问题,提出一种动态排序模式的DSSRM算法以减少差异图像分割错误,该算法建立基于合并区域的多特征马氏距离排序准则,在每次合并之后更新区域邻接矩阵并重新排序;然后,基于互信息最小化准则构造多通道差异数据集以提高算法对区域合并的约束能力;最后,提出一种级联分割变化检测框架,第1级利用SRM算法将差异图像映射到超像素空间,第2级采用DSSRM算法对超像素进行动态合并获得收敛的分割结果,第3级采用简化SRM方法进行三次合并获得最终的变化检测图。实验结果表明,该方法可以获得比SRM方法和目前流行方法更好的检测性能。
SRM (Statistical Region Merging) segmentation algorithm has the advantages of fastness, stability and anti-noise. Based on this, this paper presents a SAR image change detection method based on DSSRM (Dynamic Sorting Statistical Region Merging). Firstly, aiming at over-segmentation problem caused by SRM based on single-feature static ordering, a DSSRM algorithm of dynamic ordering model is proposed to reduce the difference of image segmentation error. The algorithm establishes a multi-feature Mahalanobis distance ordering criterion based on the merge region. After the mergence, the adjacency matrix is updated and rearranged. Then, a multichannel difference dataset is constructed based on the minimization of mutual information to improve the ability of the algorithm to bind the region. Finally, a framework of cascading segmentation change detection is proposed. In the first stage, The SRM algorithm maps the difference image into superpixels space. The second level uses DSSRM algorithm to dynamically merge the superpixels to obtain the convergent segmentation result. The third level adopts the simplified SRM method to obtain the final change detection map. The experimental results show that this method can get better detection performance than the SRM method and the current popular method.