数据与计算发展前沿 ›› 2026, Vol. 8 ›› Issue (5): 122-133.

CSTR: 32002.14.jfdc.CN10-1649/TP.2026.05.010

doi: 10.11871/jfdc.issn.2096-742X.2026.05.010

• 技术与应用 • 上一篇    下一篇

基于地形感知混合注意力网络的数字高程模型超分辨率重建

伍清林1,2(),徐瑾昊1,孙兴亮3,郭学军1,冯敏1,2,*()   

  1. 1 中国科学院青藏高原研究所， 青藏高原地球系统与资源环境重点实验室，国家青藏高原科学数据中心， 北京 100101
    2 中国科学院大学， 资源与环境学院， 北京 101408
    3 兰州大学资源环境学院， 西部环境教育部重点实验室， 甘肃 兰州 730000
  • 收稿日期:2026-03-23 出版日期:2026-10-20 发布日期:2026-10-20
  • 通讯作者: *冯敏（E-mail: mfeng@itpcas.ac.cn）
  • 作者简介:伍清林，中国科学院青藏高原研究所，硕士研究生，主要研究方向为基于深度学习的DEM超分辨率重建。
    在本文中负责模型设计与论文撰写。
    WU Qinglin is a master’s student at the Institute of Tibetan Plateau Research, Chinese Academy of Sciences. His main research area is DEM super-resolution reconstruction based on deep learning.
    In this paper, he is responsible for model design and paper writing. E-mail: qlwu@itpcas.ac.cn|冯敏，中国科学院青藏高原研究所，研究员，博士生导师，主要研究方向为青藏高原遥感大数据。
    在本文中负责论文指导与修改。
    FENG Min is a researcher and doctoral supervisor at the Institute of Tibetan Plateau Research, Chinese Academy of Sciences, whose main research area is remote sensing big data of the Tibetan Plateau.
    In this paper, he is responsible for guiding and revising the paper. E-mail: mfeng@itpcas.ac.cn
  • 基金资助:
    西藏自治区科技计划项目“青藏科考成果语料库和大语言模型建设”(XZ202601ZY0093)

A Topography-Aware Hybrid Attention Network for Building a Super-Resolution Digital Elevation Model

WU Qinglin1,2(),XU Jinhao1,SUN Xingliang3,GUO Xuejun1,FENG Min1,2,*()   

  1. 1 National Tibetan Plateau Data Center, State Key Laboratory of Tibetan Plateau Earth System, Environment and Resources, Institute of Tibetan Plateau Research, Chinese Academy of Sciences, Beijing 100101, China
    2 College of Resources and Environment, University of Chinese Academy of Sciences, Beijing 101408, China
    3 Key Laboratory of Western China’s Environmental Systems (Ministry of Education), College of Earth and Environmental Sciences, Lanzhou University, Lanzhou, Gansu 730000, China
  • Received:2026-03-23 Online:2026-10-20 Published:2026-10-20

RichHTML

1

PDF

5

摘要：

【目的】数字高程模型是地球科学研究的关键基础数据。然而现有超分辨率方法在处理全局地形结构与局部精细纹理时常面临挑战，无法兼顾全局宏观地形趋势与局部地形细节。【方法】为此，本文提出一种地形感知混合注意力网络（Topography-aware Hybrid Attention Network，THA-Net）。该模型通过一个创新的地形感知混合注意力模块（Topography-aware Hybrid Attention Block，THAB），并行融合了基于Swin Transformer的全局分支与基于卷积的局部分支，实现了对多尺度地形特征的高效协同捕捉。【结果】实验结果表明，THA-Net在喜马拉雅等复杂地区的评估中取得了最优性能。与传统双三次插值（Bicubic）和先进的EDSR、SwinIR相比，RMSE下降了1.9%~77.5%，MAE下降了1.5%~82.6%。此外，其重建的地形细节和水文网络也最接近真实地貌，为复杂地形的分析和灾害模拟提供了重要方法。【结论】全局与局部协同的DEM超分辨率重建方法能提升对地形的感知能力，从而增强超分辨率的效果。

关键词: 数字高程模型, 超分辨率重建, 深度学习, Transformer, 卷积神经网络

Abstract:

[Objective] Digital Elevation Models (DEMs) are a critical data foundation for research in the Earth sciences. However, existing super-resolution (SR) methods often face a common challenge in reconciling global topographic structures with local fine-grained textures, failing to adequately preserve both macroscopic terrain trends and fine-scale details. [Methods] To address this limitation, this paper proposes a Topography-aware Hybrid Attention Network (THA-Net). This model introduces an innovative Topography-aware Hybrid Attention Block (THAB), which concurrently integrates a Swin Transformer-based global branch with a convolution-based local branch to enable the synergistic and efficient capture of multi-scale topographic features. [Results] Experimental results demonstrate that THA-Net achieves superior performance in evaluations on complex regions, such as the Himalayas. Compared to traditional Bicubic interpolation and advanced methods like EDSR and SwinIR, THA-Net reduces the Root Mean Square Error (RMSE) by 1.9% to 77.5% and the Mean Absolute Error (MAE) by 1.5% to 82.6%. Furthermore, the reconstructed terrain details and hydrological networks from our model exhibit the highest fidelity to the ground-truth topography, offering a valuable tool for complex terrain analysis and disaster simulation. [Conclusion] We conclude that this collaborative approach, which synergizes global and local feature learning, enhances the model’s topographical awareness and leads to superior reconstruction quality.

Key words: Digital Elevation Models, super-resolution, deep learning, Transformer, Convolutional Neural Network