RESOLUTION DIFFERENCE ATTENTION NETWORK FOR CHANGE DETECTION IN CROSS-RESOLUTION REMOTE SENSING IMAGERY

In real-world satellite remote sensing applications, remote sensing images often have different spatial resolutions, which presents significant challenges for bi-temporal change detection tasks. Conventional feature extraction methods struggle to simultaneously represent the features of changed landcovers in images with different resolutions. To address these difficulties, this paper proposes a Resolution Difference Attention Network that enhances key features in images with varying resolutions, improving the model’s detection accuracy in cross-resolution scenarios. The proposed method was validated on the CDD and WHU-CD datasets, and the results demonstrate its superior performance.