EUVIMEDEUVIMED entdecken
Crossref

Efficient cloud segmentation from nychthemeron and multi-regional ground-based sky images

Ifran Rahman Nijhum, Md Shihab Reza, Soumyabrata Dev

The Visual Computer · 2026-11 · 42 (14)

DOI: 10.1007/s00371-026-04806-2 ↗

Kurz gefasst

Fachartikel in der Zeitschrift The Visual Computer, online erschienen am 06.11.2026. Verfasst von Ifran Rahman Nijhum und 2 weiteren Autor:innen. Die Arbeit nennt 44 Literaturangaben. Gefördert von: Science Foundation Ireland.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate cloud segmentation from ground-based sky images is essential for weather forecasting, solar energy prediction, and atmospheric monitoring. Cloud segmentation under both daytime and nighttime conditions remains challenging because many high-performing models rely on computationally expensive architectures that limit edge deployment. Nighttime imaging further complicates segmentation, as low illumination and sensor noise reduce feature quality. To address these challenges, we propose SEE-U-Net, an efficient encoder–decoder network built around the Squeeze–Expand–Excite (SEE) block. The SEE block combines 1×1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$1{\times }1$$\end{document} channel compression, depthwise-separable spatial expansion, and channel-wise excitation to produce discriminative features with low computational cost. We also introduce the Avro Nighttime Cloud Segmentation Dataset, a collection of nighttime sky images from Ireland, Bangladesh, and Australia for evaluating cross-region performance under low-light conditions. SEE-U-Net achieves an F1/Dice score of 0.9536 on the SWINySEG day–night benchmark, outperforming U-Net, Squeeze U-Net, SegFormer, Mobile U-Net, and DeepLabv3+. On the Avro dataset, it achieves an F1/Dice score of 0.9226 with the lowest error rate of 0.0854. Compared with standard U-Net, SEE-U-Net reduces the number of parameters by 96.7% and requires over 45×\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times $$\end{document} fewer FLOPs, providing an effective balance between segmentation accuracy and computational efficiency for real-world sky imaging applications.

Förderung

  • Science Foundation Ireland

Zitieren

APA

Nijhum, I. R., Reza, M. S., & Dev, S. (2026). Efficient cloud segmentation from nychthemeron and multi-regional ground-based sky images. The Visual Computer, 42(14). https://doi.org/10.1007/s00371-026-04806-2

Vancouver

Nijhum IR, Reza MS, Dev S. Efficient cloud segmentation from nychthemeron and multi-regional ground-based sky images. The Visual Computer. 2026;42(14). doi:10.1007/s00371-026-04806-2

BibTeX
@article{Nijhum2026,
  author = {Nijhum, Ifran Rahman and Reza, Md Shihab and Dev, Soumyabrata},
  title = {Efficient cloud segmentation from nychthemeron and multi-regional ground-based sky images},
  journal = {The Visual Computer},
  year = {2026},
  volume = {42},
  number = {14},
  doi = {10.1007/s00371-026-04806-2}
}
Kostenlos starten →