Research Progress of Transformer-based Remote Sensing Image Change Detection

Due to complex factors, including illumination, seasons, phenology, solar height and angle changes, as well as the scattered and diverse nature of the target areas to be detected, along with the variability in scale and direction, remote sensing image change detection enfaces significantly technical challenges. In recent years, Transformer has demonstrated success in various fields such as natural language processing, object detection, and image segmentation, becoming a research focus. This paper reviewed the latest research progress of Transformer-based remote sensing image change detection, and analyzed two types of methods based on pure transformer and convolutional neural network (CNN)+Transformer. Then, a comparative analysis of the detection performance of different methods on public datasets was conducted, highliting the respective merits and limitations of various methods. Finally, future possible development trends were discussed.