EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

DCA-UNet for Landslide Segmentation with Deformable Convolution and Aggregated Attention

Yingxu Song, Jie Luo, Cheng Wang, Xiangyan Kong, Yujia Zou, Yingcong Huang, Weicheng Wu, Yuan Li, Run Wang, Shiyao Li, Zuohua Tang, Shiluo Xu und 2 weitere

Remote Sensing · 16.06.2026 · 18 (12) · 2000

DOI: 10.3390/rs18122000 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Remote Sensing, online erschienen am 16.06.2026. Verfasst von Yingxu Song und 13 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Deep Earth Probe and Mineral Resources Exploration-National Science and Technology Major Project, National Institute of Natural Hazards, Ministry of Emergency Management of China, Civil Aerospace Technology Advance Research Project of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate delineation of landslide boundaries from remote sensing imagery remains challenging because landslides exhibit irregular geometry, substantial scale variation, and strong background interference. We propose DCA-UNet, a U-Net-style segmentation network that integrates deformable convolution and aggregated attention to jointly improve geometric adaptation and local-global context modeling. Deformable convolution adjusts spatial sampling locations to irregular landslide boundaries, whereas aggregated attention enhances contextual discrimination in visually ambiguous terrain. We evaluate the method on three public benchmarks—Landslide4Sense, HR-GLDD, and GDCLD—under a controlled from-scratch benchmark with dataset-specific preprocessing and official data splits. DCA-UNet achieves the best overall IoU/F1 ranking across the three datasets, reaching 61.92%/76.48% on Landslide4Sense, 59.24%/74.41% on HR-GLDD, and 58.40%/73.74% on GDCLD. The model contains 29.50 million parameters, which is close to vanilla U-Net and substantially fewer than several transformer-based baselines, although its training-side runtime and memory consumption are not the lowest. These results show that combining adaptive spatial sampling with local-global contextual aggregation is effective for landslide segmentation in both multispectral and RGB remote sensing imagery.

Förderung

  • Deep Earth Probe and Mineral Resources Exploration-National Science and Technology Major Project
  • National Institute of Natural Hazards, Ministry of Emergency Management of China
  • Civil Aerospace Technology Advance Research Project of China
  • Jiangxi Provincial Natural Science Foundation
  • Hubei Provincial Natural Science Foundation of China
  • Research Fund Program of Hubei Key Laboratory of Resources and Eco-Environment Geology
  • Geological Survey Projects of the China Geological Survey

Zitieren

APA

Song, Y., Luo, J., Wang, C., Kong, X., Zou, Y., Huang, Y., Wu, W., Li, Y., Wang, R., Li, S., Tang, Z., Xu, S., Li, Q., & Chen, H. (2026). DCA-UNet for Landslide Segmentation with Deformable Convolution and Aggregated Attention. Remote Sensing, 18(12), 2000. https://doi.org/10.3390/rs18122000

Vancouver

Song Y, Luo J, Wang C, Kong X, Zou Y, Huang Y, et al. DCA-UNet for Landslide Segmentation with Deformable Convolution and Aggregated Attention. Remote Sensing. 2026;18(12):2000. doi:10.3390/rs18122000

BibTeX
@article{Song2026,
  author = {Song, Yingxu and Luo, Jie and Wang, Cheng and Kong, Xiangyan and Zou, Yujia and Huang, Yingcong and Wu, Weicheng and Li, Yuan and Wang, Run and Li, Shiyao and Tang, Zuohua and Xu, Shiluo and Li, Qiang and Chen, Hui},
  title = {DCA-UNet for Landslide Segmentation with Deformable Convolution and Aggregated Attention},
  journal = {Remote Sensing},
  year = {2026},
  volume = {18},
  number = {12},
  pages = {2000},
  doi = {10.3390/rs18122000}
}
Kostenlos starten →