Remote Sensing Images Change Detection Based On ICA

Through change detection between multi-temporal remote sensing images of the sameregion,one can determine the change characters and process quantificationally for the remotesensing images obtained from different time.It is used for resources and environmentmonitoring,evaluation and natural disasters of battlefield analysis,etc.With development ofremote sensing,change detection has become an important research direction of remote sensingimages's application.PCA is a method based on second-order statistics,only the statistics of the signal distribution,subject to the Gaussian case,to completely eliminate the signal information,and for non-Gaussiansignal can only remove the signal between the second-order information.But in the multi-temporal remote sensing image,The spectral characteristics of various surface features do notsatisfy the Gaussian distribution,So after PCA transform,the components of the image still existsbetween high order related information,and the related information will affect changeinformation detection.Independence component Analysis(ICA) is a blind source separatetechnique appeared in recent years,It can reduce second and high-order dependences in observeddata,and the independent components are statistically as independent as possible.In thispaper,according to independent component analysis and different period of remote sensing imagecharacteristics,use independent component analysis model realize remote sensing image changedetection.Because the before and after phase's remote sensing images contains many wavebands,therefore this paper achieves the change detection based on multi-temporal remote sensingimage by using two kinds of independent component analysis.First,the independent component analysis based on principal component transform.Thismethod is to carry on the principal components transformation to Around two phase remotesensing images,make the multi-band image information into the first two principal components,then carry on the independent component analysis to two phase principal components,to achievechange detection.Second,the independent component analysis based on multi-wave band image. This methodis to select the most superior wave band combination through the analysis band spectralfeatures,information entropy,correlation coefficient of the two phase remote sensing image.Afterthe experimental study,select the band4,5,7as an experimental band combination,and then usethe three band combination of the phase two images do independent component transformation,toget change information area.Finally,use these two methods in the experiment,and compared with PCA,the results showthat the second independent component analysis method is better.