融合IR-MAD与超像素分割的遥感影像变化检测方法

Remote sensing image change detection method integrating IR-MAD and super-pixel segmentation

针对迭代加权多元变化检测(IR-MAD)方法提取高分辨率遥感影像变化结果存在信息分散、“椒盐噪声”现象,提出了一种融合IR-MAD与超像素分割的高分辨率遥感影像变化检测方法。该方法利用卡方分布概率密度函数与超像素分割方法技术对IR-MAD算法改进,通过IR-MAD算法获取初始变化检测结果;然后,采用多尺度的超像元分割方法生成多尺度超像素对象;最后,对IR-MAD算法提取的初始变化检测结果,利用多尺度超像素分割对象结合众数规则判断分割对象属性,获取最终变化区域。实验结果表明:本文提出方法能有效提取变化区域,减少“椒盐噪声”现象,能较好地保存地物边界信息,提高整体变化检测精度。

In view of the phenomenon of information dispersion and"salt and pepper noise"in the change results of high-resolution remote sensing images extracted by IR-MAD method.A change detection method of high-resolution remote sensing image integrating IR-MAD and super-pixel segmentation was proposed.In this method,the chi square distribution probability density function and super-pixel segmentation method were applied to improve the IR-MAD algorithm,and the initial change detection results were obtained by the IR-MAD algorithm;Then,the multi-scale super-pixel object was generated by multi-scale super-pixel segmentation method;Finally,the initial change detection results extracted by IR-MAD algorithm were obtained by multi-scale super-pixel segmentation object,combined with mode rules to obtain the final change region.The experimental results showed that the proposed method could effectively extract the change area,reduce the phenomenon of"salt and pepper noise",better preserve the boundary information of ground objects and improve the overall change detection accuracy.