Related Experiment Video
Updated: Jun 7, 2025
10:35
Bringing the Visible Universe into Focus with Robo-AO
Published on: February 12, 2013
19.4K
A model for suppressing stray light in astronomical images based on deep learning.
Mo Chen1, Yan Zhao2, Wenbo Yang3
1College of Communication Engineering, Jilin University, Nanhu Road No.5372, Changchun, 130012, Jilin, China.
Scientific Reports
|November 11, 2024
Summary
A new deep learning model, the Pyramid Deformable Large Kernel Attention (PD-LKA) Model, effectively suppresses stray light in astronomical images. This enhances observation quality and improves object positioning accuracy for Wide-Field Small Aperture Telescopes (WFSAT).
Area of Science:
- Astronomy
- Deep Learning
- Image Processing
Background:
- Wide-Field Small Aperture Telescopes (WFSAT) are crucial for space object surveillance due to their wide fields of view.
- Stray light, from sources like moonlight and clouds, significantly degrades WFSAT image quality and can impair object positioning.
Purpose of the Study:
- To develop a deep learning model for effective stray light suppression in astronomical images.
- To improve the robustness and accuracy of WFSAT observations under stray light conditions.
Main Methods:
- Proposed the Pyramid Deformable Large Kernel Attention (PD-LKA) Model, incorporating a pyramid structure for multi-scale feature capture and Deformable Large Kernel Attention (D-LKA) for precise interference localization.
- Trained the model using simulated astronomical image pairs and validated its performance on real-world image sequences.
Main Results:
- Achieved high performance metrics on simulated data, with a Peak Signal-to-Noise Ratio (PSNR) of 32.540 and a Structural Similarity Index Measure (SSIM) of 0.938.
- Demonstrated improved object positioning accuracy, better than 5 arcseconds, on real stray light-affected image sequences.
Conclusions:
- The PD-LKA Model successfully suppresses stray light while preserving essential image characteristics such as gray values, shapes, and positional information.
- The model enhances the reliability and accuracy of astronomical observations made with WFSATs, even in the presence of significant stray light interference.
