LDG-Net: a lightweight dual-gating network for sea-land segmentation through multimodal data fusion using SAR and optical imagery

Accurate sea-land segmentation is essential for coastal applications such as shoreline change detection and marine spatial planning. However, single-modality methods often fail to capture complete surface information under complex environmental conditions. Meanwhile, many existing multimodal approaches rely on computationally intensive architectures, limiting their suitability for real-time or resource-constrained scenarios. To address these challenges, a lightweight dual-gating network, LDG-Net, is proposed for accurate and efficient sea-land segmentation. LDG-Net exploits optical and SAR imagery to achieve robust sea-land segmentation. LDG-Net proposes a novel hierarchical spatial-channel dual-gating mechanism through our newly-designed Spatial-Gated Modality Selection (SGMS) and Channel-Gated Modality Fusion (CGMF) modules to mitigate spatial heterogeneity and cross-modal feature discrepancies. Furthermore, by employing an EfficientNet encoder with depthwise separable convolutions in the decoder, the network achieves enhanced segmentation accuracy while maintaining computational efficiency. Experimental results demonstrate that LDG-Net outperforms state-of-the-art methods in both accuracy and efficiency, offering a practical solution for complex coastal environments.