Change detection based on image segment and fusion in multitemporal SAR images

Abstract

In this paper, a new change detection method based on image segmentation and fusion in multi-temporal SAR image is presented. The proposed fusion method exploits the conditional probability of difference image to two original images to fuse the segment results of two original SAR images. The new method can be divided into four steps: (1) the original SAR images are segmented into two classes using the proposed segmentation method based on canny edge match. (2) The two segmentation result images are fused with conditional probability rule and a fusion difference image considering spatial context is obtained. (3) The fusion image is thresholded with OTSU threshold, and the initial change detection result is obtained. (4) With subtraction operation and absolute operation between two segment results of original SAR images, a Boolean image with 0 and 1 value is obtained. Then the change detection result is composed of the regions in the Boolean image corresponding to the change regions in the initial change detection result. Experimental results obtained on simulate and real SAR images confirm the effectiveness of the proposed approach.