Answers are grounded in this paper and cite their sources.
Remote sensing-based fine rice mapping is essential for crop acreage statistics, yield estimation, food security assessment, and precision agriculture. However, accurate rice segmentation remains challenging because rice fields are often fragmented, field boundaries are narrow and irregular, and rice can show high spectral and visual similarity to other vegetation or water-related backgrounds. Single-source imagery is usually insufficient to address these issues. High-resolution imagery (HRI) preserves field texture and boundary details but lacks fine spectral discrimination, whereas hyperspectral imagery (HSI) provides crop-sensitive spectral responses but suffers from lower spatial resolution. To exploit the complementary advantages of both sources, this article proposes DGSC-MambaFuse, an HRI–HSI multisource remote sensing fusion network based on Mamba sequence modeling for fine rice mapping. The network contains an HRI spatial encoder, a rice-sensitive HSI encoder, a Derivative-Guided Spectral Calibration (DGSC) module, a Cross-Source Gated Fusion Mamba (CSGF-Mamba) module, and a refinement decoder. DGSC enhances rice-sensitive spectral representations through band-level calibration, local spectral modeling, and spectral derivative calibration. CSGF-Mamba then adaptively exchanges spatial and spectral information at deep semantic levels, reducing the adverse effect of cross-source resolution mismatch while preserving fine field structures. Experiments were conducted on a dual-source rice segmentation dataset constructed from Jilin-1 HRI and Zhuhai-1 HSI in Anxiang County, Hunan Province, China. The proposed method reached 92.84% OA and 85.31% mIoU. Compared with the strongest single-source HRI and HSI baselines selected by mIoU, SAM2-UNeXT-HRI, and AFENet-HSI, it improved OA and mIoU by 3.22 and 6.11 percentage points and by 3.06 and 5.90 percentage points, respectively. It also outperformed the evaluated multisource baselines.
No takes yet. Share an insight, caveat, or question.
Yang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: