基于KICA的多时相遥感图像变化检测

摘要

针对PCA变化检测方法的精度较低和ICA方法的线性局限性问题，提出了基于核独立成分分析( KI-CA)的多时相遥感图像变化检测方法。首先，将每一时相的图像转化为列向量，并把这些列向量组成矩阵；然后，通过核函数将矩阵映射到高维特征空间中，再在该空间中利用ICA方法分离出相互独立的图像分量；接着通过FCM算法分割表征变化信息的图像分量，并采用区域生长算法获得完整的变化信息；最后，分别利用本文方法与差值法、PCA方法和ICA方法对多时相遥感图像进行变化检测，并对检测结果进行定性分析和定量比较。结果表明，该方法能更好地分离出多时相遥感图像的变化信息，具有更高检测精度。%Aiming at the problem of low accuracy of the PCA change detection method and linear limitation of the ICA, a change detection method of multi-temporal remote sensing images based on kernel independent component analysis ( KICA) is proposed. Firstly, each temporal remote sensing image is reshaped into a column vector, and these vectors are used to form a matrix. Then the matrix is mapped into a high dimensional feature space by the kernel function. And in the feature space mutually independent image components are separated by the ICA meth-od. Next, the FCM algorithm is used to segment the image component representing changed information, and the region growing algorithm is used to get complete changed information. Finally, change detection experiments on multi-temporal remote sensing images are carried out by the proposed method, the image difference method, PCA method and the ICA method, respectively. And some qualitative analysis and quantitative comparisons of change detection results are given. The experimental results show that the proposed method can separate the changed infor-mation of multi-temporal remote sensing images better and has higher accuracy.