基于小波变换和核独立分量分析的遥感图像变化检测

Remote Sensing Image Change Detection Based on Wavelet and Kernel Independent Component Analysis

为了进一步提高基于独立分量分析(Independent Component Analysis,ICA)的遥感图像变化检测精确度,并解决ICA分离的图像分量排序不确定问题,提出了基于小波变换和核独立分量分析(Kernel Independent Component Analysis,KICA)的遥感图像变化检测方法。首先通过小波变换对遥感图像进行分解,得到由图像的高频分量和低频分量组成的分块向量,然后利用核函数将分块向量映射到高维特征空间中,再在该空间中用ICA方法分离出互相独立的向量,最后根据分离出的向量中高频分量的差异自动分辨出变化分量。文章给出了遥感图像变化检测方法及近年提出的基于主分量分析(Principal Component Analysis,PCA)、基于ICA、基于KICA三种变化检测方法的试验结果,并进行了分析和定量比较。试验结果表明,文中方法能更好地分离出遥感图像的变化信息,具有更高的精确度,并实现了变化检测的智能化。

In order to further improve the accuracy of change detection of remote sensing images based on independent component analysis （ICA） and to solve the uncertainty problem of sequence of image components separated by ICA, a change detection method based on wavelet transform and kernel independent component analysis （KICA） was proposed. Firstly, the remote sensing images were decomposed by wavelet transform, and partitioned vectors composed of high-frequency components and low-frequency components were obtained. Then the partitioned vectors were mapped into a high-dimensional feature space by the kernel function, and the mutually independent vectors were separated by ICA in this space. Finally, according to the differences between the high-frequency components of the separated vectors, the change component was distinguished automatically. Experimental results of the proposed method and other three change detection methods proposed recently based on the principal component analysis （PCA）, ICA, KICA were given. And some analysis and quantitative comparisons were done. A large number of experimental results show that the proposed method can separate change information of remote sensing images with higher accuracy, and the intelligent change detection is realized.