Unsupervised SAR change detection based on a new statistical model

In this paper, with the help of a new statistical model, the problem of detecting the changes that occurred on the ground by analyzing SAR imagery is addressed by a completely unsupervised approach. In the proposal change detection approach, a difference image of two multitemporal SAR images is firstly produced by the generalized likelihood ratio operator. After that, a new statistical model is particularly developed to model the distribution of generalized likelihood ratio test. At last, With the help of the new statistical model, an automatic thresholding procedure is performed on the difference image to detect changes. Experimental results obtained on real SAR images acquired by the CARABAS-II confirm the effectiveness of the change detection approach.