Destructive disasters can cause enormous harm and losses, and due to a lack of information for a certain period of time after the disaster, it is extremely detrimental to understanding the disaster situation and disaster relief. In order to obtain timely collapse and damage information of buildings, roads, bridges, reservoirs and other important ground objects in disaster areas, a method for automatically identifying and annotating the differences in remote sensing images before and after disasters has been proposed. Firstly, Gaussian noise was removed from time-series remote sensing images using block matching 3D (BM3D) method, and then image registration was performed using scale invariant feature transform (SIFT) method. The edge information of important ground objects in different areas was obtained by using Wv_Canny edge detection method for difference images. Finally, the important ground objects that have changed were identified and annotated. The experimental results of remote sensing image were compared according to the change area of the building, the correct rate is 78% ~ 79%, the false detection rate is 21% ~ 22%, and there is no missed detection rate. Simulation experiment and actual remote sensing image processing show that the proposed method can effectively identify and mark the difference areas of important ground objects such as buildings, which is conducive to timely understanding and rescue of destructive ground objects after disaster. [ABSTRACT FROM AUTHOR]