规整化策略的面向对象变化检测研究

为了提高高分辨率遥感影像变化检测的精度,该文提出一种基于规整化策略的面向对象迭代加权多变量变化检测算法。该方法利用多尺度分割法对两期影像进行了分割并提取了影像对象的各种特征,选择具有代表性的特征参与面向对象的IR-MAD变化检测,并在迭代加权的过程中加入规整化策略,避免广义特征方程可能出现的不稳定性。该方法减少了噪声,提取了研究区大部分变化区域,提高了高分辨率影像的变化检测精度和可靠性。结合人工变化检测和像素级IR-MAD检测结果,并采用新疆边界口岸资源三号卫星影像,验证了该方法的有效性。

Object-oriented IR-MAD change detection based on normalization strategy

In order to improve the accuracy of high-resolution remote sensing image change detection,an object-oriented iterative weighted multivariate change detection algorithm based on normalization strategy is proposed.In this method,multi-scale segmentation method is used to segment two images,and each feature of image object is extracted.The representative feature is selected to participate in the IR-MAD change detection of object-oriented,and the regularization strategy is added to the iterative weighting process to avoid the instability of the generalized characteristic equation.This method reduces the noise,extracts most of the change area of the research area,and improves the accuracy and reliability of the change detection of high resolution image.Combined with the artificial change detection and pixel level IR-mad detection results,the effectiveness of the method was verified by using ZY-3 satellite image.