The paper presents a tool,the morphological model (MP),is used in sy nthetic aperture radar (SAR) images,giving shape to the multi-dimensional details of the SAR image,improving the detection accurac y of the image.Adding algorithms to image detection can be applied to remote sensing image change detection.SAR image change detect ion method is proposed in this paper the multi-scale fusion spectral subspace clustering,on the basis of this method is with figure on the data obtained from different time change the logarithmic ratio and average ratio processing,thus forming the structure chara cteristics of the multiple word space image,combined with the random sampling of spectral clustering subspace fusion algorithm to get the change of the image.The SAR image obtained by this method can detect many single pixels,and the number of false detected pixe ls is significantly reduced,which is better than other algorithms.The detection results suppress the missing problem in the image chan ge,so the detection accuracy is high.