Research Groups
Key Laboratories
Research Progress
Research Programs
Quick Links
Location: Home > Research > Research Progress
Researchers Make Progress in Identification of Contaminated Images in Light Curve Data Preprocessing
Author: | Update time:2024-05-30           | Print | Close | Text Size: A A A

The Applied Astronomy Group at Yunnan Observatories has made progress in identification of contaminated images in light curve data preprocessing. The researchers have developed, trained, and validated machine learning models to detect and classify images contaminated by stars and clouds. The results have been published in the journal Research in Astronomy and Astrophysics, under the title “Machine Learning-Based Identification of Contaminated Images in Light Curves Data Preprocessing”.

Statistical analysis has shown that contamination can be divided into two main categories: the first type, "stellar contamination," occurs when nearby stars cause a significant increase in the target object's brightness; the second type, "cloud contamination," results from cloud cover causing a noticeable decrease in brightness. Traditionally, manual inspection of images is time-consuming and labor-intensive. Therefore, we propose the use of machine learning methods for image classification. Convolutional Neural Networks (CNN) and Support Vector Machines (SVM) are used to detect stellar contamination and cloud contamination, respectively, achieving F1 scores of 1.00 and 0.98 on the test set.

A dataset for machine learning training has been created through manual annotation of partial observation data from the 1.2m telescope at Yunnan Observatories for the years 2022-2023. For identifying stellar contamination, the CNN model trained on stamp images demonstrates higher accuracy. For identifying cloud contamination, the ResNet-18 model shows low accuracy on both training and test sets; the lightGBM model has poor generalization performance, requiring long-term data accumulation to overcome this issue; the SVM model exhibits strong generalization capabilities, maintaining a high classification accuracy of 97.12% for new samples.

This work was funded by the National Natural Science Foundation of China (NSFC, Nos. 12373086 and 12303082), CAS “Light of West China” Program, Yunnan Revitalization Talent Support Program in Yunnan Province.

LI Hui
Yunnan Observatories, CAS

Copyright © 2013 Yunnan Observatories, CAS All Rights Reserved.
Address: 396 Yangfangwang, Guandu District, Kunming 650216, Yunnan, China
Tel: +86 871 63920919 Fax: +86 871 63920599