LGMM-Net: A Local–Global Encoder and Mask Mamba Decoder Network for Remote Sensing Change Detection

Remote sensing change detection (CD) is critical for applications such as urban planning and disaster assessment. It enables accurate identification of changes in ground objects across bi-temporal images. With the rapid development of deep learning, hybrid Convolutional Neural Network (CNN)–Transformer approaches have significantly advanced this field. However, existing fusion networks often lack robustness. They frequently misclassify pseudo-changes as true ones. Even with hard channel-splitting strategies that jointly model local and global features, this issue remains unresolved. To address these challenges, we propose a novel Local-Global Encoder and Mask Mamba Decoder Network (LGMM-Net). The network employs multi-kernel convolutions, window attention, and state-space sequence modeling to achieve robust fusion of local details and global semantics. First, we introduce a dynamic channel soft-splitting strategy within the Dynamic Local–Global Parallel Encoder (DLGPE). This encoder integrates multi-kernel convolutions, fixed-window attention, and shifted-window attention to capture both local details and global semantics. Second, we design the Pyramid Mask Mamba (PMM) module with a new scanning mechanism, termed Mask-Guided Directional Selective Scan (MGD-SS2D). In this module, directional convolutions define scanning paths, while feature rotations enable four-directional modeling, thus enhancing boundary representation. Extensive experiments on four public datasets demonstrate that LGMM-Net consistently outperforms state-of-the-art (SOTA) methods in both accuracy and robustness. Our source code is available on https://github.com/Jhin1215/LGMM.