Related Experiment Video
Updated: Jul 3, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
Deep learning-based semantic segmentation of night-sky clouds for operational telescope scheduling
Xuan Liu1, Hai Cao2,3, Ruojun Wang4
1School of Airspace Science and Engineering, Shandong University, Weihai, 264209, Shandong, People's Republic of China.
Scientific Reports
|July 1, 2026
Summary
Astronomical observatories can now improve efficiency with NightCloudSegNet, a new deep learning model for real-time cloud detection. This system provides crucial dome-scale cloud coverage data, optimizing telescope scheduling and preventing data loss.
Area of Science:
- Astronomy
- Computer Vision
- Machine Learning
Background:
- Ground-based optical telescopes require precise, real-time cloud data for efficient operation.
- Inaccurate or delayed atmospheric information leads to wasted observation time and lost scientific data.
Purpose of the Study:
- To introduce the WOANC dataset and NightCloudSegNet, a novel framework for cloud segmentation in astronomical imaging.
- To enhance automated telescope scheduling and shuttering decisions through accurate cloud coverage assessment.
Main Methods:
- Developed NightCloudSegNet, a fisheye-aware segmentation framework tailored for low-light conditions.
- Utilized the WOANC dataset, a pixel-annotated nighttime full-dome dataset from an operational observatory.
- Evaluated model performance on both the WOANC and external SWINSEG datasets.
Main Results:
- NightCloudSegNet achieved 86.6% mIoU and 92.8% F1 score on the WOANC test set.
- The model demonstrated robust performance on the SWINSEG dataset with 86.2% mIoU and 92.6% F1 score.
- Successfully translated segmentation masks into per-target observability indicators.
Conclusions:
- NightCloudSegNet offers a reliable solution for real-time cloud detection in astronomical observatories.
- The framework's performance indicates its potential to significantly improve observational efficiency in automated telescope operations.
- Accurate cloud segmentation supports informed decisions, maximizing the use of valuable telescope resources.
