Change Detection of Multi-spectral Remote Sensed Images Based on PCA and EM Algorithm

Change detection can get useful changed information from the observed data about the earth.In this paper,a new technology aimed at two aspects in the directly compared way is based on pixel’s characters in the multi-spectral and multi-temporal remote sensed images,which are "build the different image" and "threshold determination".In our method,an exact difference image can be made by PCA method,and then the threshold from the EM algorithm is used to obtain changed area,which can be seen as the result.Compared with the traditional method,the proposed method has a higher accuracy,and is more effective.