迭代加权多元变化检测算法在高分辨率遥感影像变化检测中应用  

Application of iteratively re-weighted multivariate alteration detection in high-resolution remote sensing images change detection

作　　者：张续 江涛[2] 胡世明 焦帅[1] ZHANG Xu;JIANG Tao;HU Shiming;JIAO Shuai(College of Resource Environment and Tourism. Capital Normal University, Beijing 100048, China;College of Geomatics, Shangdong University of Science and Technology, Qingdao Shandong 266590, China)

机构地区：[1]首都师范大学资源环境与旅游学院,北京100048 [2]山东科技大学测绘科学与工程学院,山东青岛266590

出　　处：《计算机应用》2019年第A01期177-181,共5页journal of Computer Applications

基　　金：国家自然科学基金资助项目(41706194);山东省自然科学基金资助项目(ZR2016DB23);科技创新服务能力建设-基本科研业务费(科研类)资助项目(025185305000/191)

摘　　要：针对传统变化检测算法在高分辨率遥感影像变化检测中精度低的问题,探讨了迭代加权多元变化检测(IR-MAD)算法在高分辨率遥感影像变化检测中的有效性。首先通过多元变化检测(MAD)算法计算原始影像的特征值和特征向量,获得典型变量;然后利用最大期望算法(EM)计算典型变量的先验概率、均值和标准差;最后通过迭代加权获得每个像元稳定不变的权重,通过权重与阈值的比较判断像元是否发生变化,提取变化区域。使用高分辨率遥感影像WorldView和QuickBird数据作为实验数据,并与传统的变化向量分析法和主成分分析K均值法进行对比分析,结果表明,IR-MAD算法在高分辨率遥感影的变化检测中具有优势。Aiming at the low accuracy of traditional change detection methods in detecting high-resolution remote sensing images, the effectiveness of Iteratively Re-weighted Multivariate Alteration Detection (IR-MAD) in detecting the change of high-resolution remote sensing images was discussed. Firstly, Multivariate Alteration Detection (MAD) was used to calculate the eigenvalue and eigenvector to get the canonical variable. Then Expectation Maximization (EM) was used to calculate the prior probability, means and standard deviation of the canonical variable.Finally, the stable weight of each pixel was obtained by iterative re-weighting and was compared with the threshold value to judge whether the pixel changed, and then the changed area was extracted. With high-resolution remote sensing images WorldView and QuickBird used as experimental data, IR-MAD shows advantages in change detection of high-resolution remote sensing images compared with Pixel-based Change Vector Analysis (PCVA) and Principal Component Analysis (PCA) K -means.

关 键 词：变化检测 像元级 高分辨率 迭代加权多元变化检测 

分 类 号：TP237[自动化与计算机技术—检测技术与自动化装置]