融合差异图与高斯混合模型相结合的SAR图像变化检测  

Change Detection in Synthetic Aperture Radar Images Based on Image Fusion and Gaussian Mixture Model

作　　者：高新 靳国旺[1] 熊新[1] 徐娇 GAO Xin;JIN Guowang;XIONG Xin;XU Jiao(Information Engineering University, Zhengzhou 450001, China)

机构地区：[1]信息工程大学,河南郑州450001

出　　处：《测绘科学技术学报》2020年第1期68-73,共6页Journal of Geomatics Science and Technology

基　　金：国家自然科学基金项目(41474010,61401509)。

摘　　要：在SAR图像变化检测过程中,假设差异图类别服从单一分布,构造单一类型的差异图不能完好地保留变化信息,从而影响SAR图像变化检测的精度。针对上述问题,提出一种基于差异图融合和高斯混合模型GMM(Gaussian Mixture Model)的非监督SAR图像变化检测方法。该方法利用给定的融合因子进行差值和比值图融合,采用最大期望算法EM(Expection Maximum)求解融合差异图的GMM参数,并根据后验概率将图像像素分配到GMM各个分量,从而获得SAR实验区域的变化检测图。3组SAR数据集的变化检测实验验证了所提方法的可行性和有效性。In the process of SAR image change detection,it can not preserve the integrity of the change information assuming the difference map subjecting to a single distribution and constructing a single type of difference map,which affects the accuracy of SAR image change detection.To solve the problem,an unsupervised SAR image change detection method is proposed based on difference graph fusion and Gaussian mixture model.In the algorithm,a given factor is used to fuse the difference graph and the ratio graph,then the EM(Expection Maximum)algorithm is used to solve the GMM(Gaussian Mixture Model)parameters of the fused difference graph.And the image pixels are assigned to the components of the GMM according to the posterior probability.Thereby,a map of the change detection results is obtained.Experiments carried out on three sets of SAR datasets show that the proposed method is feasible and effective.

关 键 词：变化检测 SAR图像 差异图融合 EM算法 高斯混合模型 

分 类 号：P237[天文地球—摄影测量与遥感]