通过对传统变化检测方法存在不足的分析 ,引进典型相关分析的基础理论 ,将不同时相的多通道遥感数据视为分组的多元随机变量 ,利用典型变换进行遥感数据的多元变化检测。实验结果表明 ,所提出的M变换方法用于多时相、多通道遥感影像的变化检测具有明显的优势和应用前景。 By analyzing the traditional detection methods, this paper introduces the basic theory of Canonical Correlation Analysis (PCA), and considers the multi-channel remote sensing data with different phases as the multivariate random variables of the group. Multivariate detection of the remote sensing data is carried out by using typical transformation. Experimental results show that the proposed M transform method has obvious advantages and application prospects for the detection of multi-temporal and multi-channel remote sensing image changes.