Machine Learning-Based Identification of Contaminated Images in Light Curves Data Preprocessing

Aug 4, 2026

Upload by Aria

Attitude estimation of space objects relies on analyzing light curves, but photometric observations often contain outliers caused by different contamination mechanisms. This work separates two dominant categories: stellar contamination from nearby stars increasing brightness, and cloudy contamination from cloud cover decreasing brightness. Manual inspection is labor-intensive, so machine learning is used to automatically identify contaminated images. Convolutional Neural Networks and Support Vector Machines achieve F1 of 1.00 for stellar and 0.98 for cloudy contamination on the test set. Additional models, including ResNet-18 and lightGBM, are compared to evaluate performance differences.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Abstract

Introduction

Data

Telescope and observations

Image dataset

Research in Astronomy and Astrophysics manuscript no.(LATEX: paper.tex; printed on April 3, 2024; 1:51) arXiv :2404 .01691v1 [ astro-ph .IM] 2 Apr 2024 Machine Learning-Based Identification of Contaminated Images in Light Curves Data Preprocessing Hui Li 1 ,2 , Rong-Wang Li 1 ,3 , Peng Shu 1 and Yu-Qiang Li 1 ,3 1 Yunnan Observatories, Chinese Academy of Sciences, Kunming 650216, China; [lihui@ynao.ac.cn](lihui@ynao.ac.cn) 2 University of Chinese Academy of Sciences, Beijing 100049, China 3 Key Laboratory of Space Object and Debris Observation,Chinese Academy of Sciences,Nanjing 210023, China Abstract Attitude is one of the crucial parameters for space objects and plays a vital role in collision prediction and debris removal. Analyzing light curves to determine attitude is the most commonly used method. In photometric observations, outliers may exist in the obtained light curves due to various reasons. Therefore, preprocessing is required to remove these outliers to obtain high quality light curves. Through statistical analysis, the reasons leading to outliers can be categorized into two main types: first, the brightness of the object significantly increases due to the passage of a star nearby, referred to as “stellar contamination,” and second, the brightness markedly decreases due to cloudy cover, referred to as“cloudy contamination.” Traditional approach of manually inspecting images for contamination is time-consuming and labor-intensive. However, We propose the utilization of machine learning methods as a substitute. Convolutional Neural Networks (CNN) and Support Vector Machines (SVM) are employed to identify cases of stellar contamination and cloudy contamination, achieving F1 scores of 1.00 and 0.98 on test set, respectively. We also explored other machine learning methods such as Residual Network-18 (ResNet-18) and Light Gradient Boosting Machine (lightGBM), then conducted comparative analyses of the results. Key words: techniques: image processing—methods: data analysis—light pollution 1 INTRODUCTION Light curves refers to the curves depicting changes in luminosity. The photometry of space objects is influenced by various factors, including the geometry involving the Sun, the space object, and the observer, as well as the object’s shape, orientation, and surface reflectance characteristics. Light curves are essential for studying the rotation state and characteristics of space objects. However, before obtaining light curves, it is necessary to preprocess the source data, which includes the removal of outliers and data contaminated by 2 H. Li, R.-W. Li, P. Shu, & Y.-Q. Li The usual preprocessing method often requires manual judgment. But when dealing with large volume of data, this judgment is time-consuming and labor-intensive. Using machine learning for pattern recognition can significantly improve efficiency and save a substantial amount of time and effort. Machine learning has widespread applications in Astronomy, including but not limited to predicting atmospheric seeing in optical observations (Ni et al. 2022), identifying AGN and pulsar candidates (Zhu et al. 2021), detecting outliers in astronomical images (Han et al. 2022), and classifing Gaia data (Bai et al. 2018) . Hinton & Salakhutdinov (2006) published a paper with two main points: (1) Artificial neural networks with multiple hidden layers exhibit exceptional feature learning capabilities. (2) The effective overcoming of training difficulties in deep neural networks can be achieved through “layerwise pre-training,” which introduced the field of deep learning (Zhou et al. 2017) . In fact, there were even highly efficient deep learning models proposed before 2006, such as CNN. In the 1980s and 1990s, some researchers published studies on CNN in the field of pattern recognition, showing excellent performance in handwritten digit recognition (Lawrence et al. 1997, Neubauer 1998) . However, at that time, CNN still performed poorly wi

FAQs

What two main types of contamination are considered in the preprocessing of light curve data?

The study categorizes outlier sources into stellar contamination, where brightness increases due to a nearby star, and cloudy contamination, where brightness decreases due to cloud cover.

Why is machine learning used instead of manual inspection?

Manual inspection of contaminated images is time-consuming and labor-intensive when processing large volumes of data, so machine learning improves efficiency and reduces workload.

Which models are used to identify stellar contamination and cloudy contamination, and how well do they perform?

A CNN is used for binary classification of stellar contamination, while CNN, lightGBM, SVM, and ResNet-18 are used for cloudy contamination classification. Reported F1 scores are 1.00 for stellar contamination and 0.98 for cloudy contamination on the test set.

Aug 4, 2026

Upload by Aria