基于聚类分析的SAR图像变化检测

针对传统变化检测方法用于合成孔径雷达(SAR)图像时不能有效地降低背景变化导致的检测虚警,提出了基于聚类分析的SAR图像变化检测方法。该方法对聚类后SAR图像像素类间变化引入Mahal-anobis距离(M距离),结合变化阈值的选择来分析图像中变化像素的M图,进而实现SAR图像变化检测。仿真结果说明:该方法不仅能克服自然和背景变化给检测带来的困难,而且,不受传感器位置的影响,能有效地对不同时刻的SAR图像进行变化检测。

Detection of target change in SAR images based on cluster analysis

According to the deficiency of the traditional change detection methods used to SAR images,which can not decrease the false alarm caused by the background changes efficiently,a cluster-based approach for detecting changes in SAR images is proposed.This method detects the changes in SAR images by introducing Mahalanobis distance(M-distance)to the different clustered images class,and analyzing the changes of M-distance of the pixels by combining the selection of the change threshold value.The simulation results show this method not only can overcome the negative affects caused by changes of natural environment and background,but also is not influenced by the position of the sensors,and can detect the changes in SAR images at different times efficiently.