Dual-Branch Cross-Resolution Interaction Learning Network for Change Detection at Different Resolutions

Change detection (CD) plays a critical role in remote sensing (RS) image analysis. However, the detection accuracy is often compromised due to differences in imaging conditions between bitemporal images, especially in scenarios where images have varying spatial resolutions from multisource RS satellites. To address this challenge, we propose an innovative dual-branch cross-resolution interaction learning network (DCILNet). This network employs strategies for image spatial resolution alignment and feature space correction to achieve efficient CD. To fully leverage the information from images with different resolutions, we design two cross-resolution CD branches. These branches interact through a cross-resolution feature correction module (CRFCM) and a multiresolution feature fusion module (MRFM), facilitating feature interaction and learning across branches to maximize feature representation. We conduct both qualitative and quantitative experiments on three public datasets. The experimental results show that, compared to other comparative methods, the proposed DCILNet exhibits stronger competitiveness. Our research shows that by integrating image spatial resolution alignment and feature space correction strategies and adopting dual-branch interactive learning, the model effectively addresses the challenges posed by resolution discrepancies in CD tasks. Our code will be available at https://github.com/Li738/DCILNet.