利用小波域HMC模型进行遥感图像变化检测

Change detection in multi-temporal remote sensing images based on the wavelet-domain hidden Markov chain model

传统阈值检测算法都是基于单函数模型进行的,当差异影像分布函数较复杂时检测结果较差.针对这个问题,提出一种基于小波域的隐马尔科夫链模型的遥感图像变化检测算法.将双高斯混合模型与小波变换结合,解决了单函数模型匹配率低的问题,并通过小波变换引入了图像的空间信息,提高了检测精度.利用双高斯混合模型对小波分解后的多层差异影像进行拟合,根据拟合结果判定待检测点类别.对得到的多层初始分割结果,利用隐马尔科夫链模型根据连续最大后验概率融合,得到最终变化检测图.对真实遥感数据集进行实验,证明这种算法可以得到较好的检测结果.

The traditional threshold algorithms detect the changes in multitemporal remote sensing images based on the analysis of the signal function model, which has a poor accuracy for difference images with complex distribution. In this paper, a new approach is proposed by virtue of the double Gaussian mixture model and the wavelet transform. The proposed algorithm has better matching than the signal function model and introduces the spatial information by using the wavelet transform. After using the double Gaussian mixture models to detect the changed regions, the change maps in different scales are fused using the HMC model based on sequential maximum a posteriori estimation. The experiments on the real remote sensing images confirm the effectiveness of the proposed algorithm.