中文摘要：

针对传统变化检测方法应用于高分辨率遥感影像变化检测时出现的变化信息分散、椒盐噪声影响严重等问题,将面向对象技术和迭代加权多变量变化检测方法结合起来,提出了一种面向对象的迭代加权多变量变化检测方法（iteratively reweighted multivariate alternative detection,IR-MAD）。该方法主要通过结合卡方分布的概率密度函数和面向对象技术来对传统多变量变化识别方法（multivariate alternative detection,MAD）进行改进,卡方分布的概率密度函数对变化信息进行融合以获取信息集中的IR-MAD变量。此外,在对影像进行分割时结合叠置分割技术获取边界一致、同质性较好的影像对象。实验表明,面向对象IR-MAD方法能够有效集成变化信息,准确提取变化区域,同时较好地保持变化目标的结构与形状,减少椒盐噪声的影响,检测结果具有较高的可靠性。

英文摘要：

Against the scatter of the change information and the effect of the salt and pepper noise when the traditional change detection method is applied to the high resolution remote sensing image change detection, by mean of the object-oriented technique and the iteratively weighted multivariate alternative detection methods, we propose an object-oriented iteratively reweighted multivariate alternative detection, IR-MAD. The method is mainly applied through the combination of the probability density function of the Chi square distribution and object-oriented technology to improve the traditional multivariate alternative detection, MAD. The probability density function of the Chi square distribution are fused with the MAD variables so as to obtain the IR-MAD variables that have the more concentrated change information. In addition,in the image segmentation,it obtains the image object with the consistency boundary and good homogeneity by combining overlapping segmentation techniques. The test results show that the object based /R-MAD method can effectively integrate change information, accurately extract the change regions, retain the better structure and shape of the changing targets and reduce the impact of the salt and pepper noise, while the detected results have high reliability.