Abstract: According to an analysis of support vector machine(SVM) and multiple kernel theory,an improved SVM change detection model based on multi-feature differencing kernel for remote sensing imagery was proposed.The combinations of kernel functions using spectral data and textural feature were discussed.After detailing the structure of the image differencing kernel based on multi-features,the algorithm of SVM change detection model was designed,and combined with category weights for the extraction of the spatial distribution of several change classes.Experimental results show that with the help of a multiple kernel function,the change detection model can get higher detection accuracy than the traditional methods,and also avoid determining change thresholds that are complex and uncertain.