Change Detection of Building Remote Sensing Images Based on RMT-BIT

The detection of building changes through remote sensing image allows for the timely monitoring of urban development dynamics, providing a scientific basis for urban planning and management. Given the complexity of spatial features in remote sensing images of buildings, this paper introduces spatial prior knowledge in terms of both space and distance. Building features from remote sensing images are extracted using retentive networks meet vision transformers (RMT). The extracted feature maps from two temporal phases are processed by a bitemporal image transformer (BIT), leading to the output of the decoder, from which the areas of change between the two phases are computed. The proposed method is validated on the WHU-CD building change detection dataset, and the results show that the proposed method effectively improves the accuracy of building change detection in remote sensing images.