Sky Seeing Estimation Using Nonparametric Fuzzy System of Low-Quality All-Sky Camera Images

• Published in: IEEE transactions on instrumentation and measurement 2024, Vol.73, p.1-17
• Main Authors: Helmy, Islam, Shokry, Ahmed, Eid, Doaa, Choi, Wooyeol
• Format: Article
• Language: English
• Subjects: All-sky camera; Astronomical models; Atmospheric modeling; Brightness; Cameras; Celestial bodies; Extraterrestrial measurements; full-width at half-maximum (FWHM); Fuzzy logic; Fuzzy sets; Ground-based observation; Image acquisition; Image contrast; Image enhancement; Image quality; Image resolution; Light pollution; Machine learning; Night sky; Observatories; Pollution measurement; seeing estimation; Sky brightness; Stars; Telescopes
• ISSN: 0018-9456; 1557-9662
• DOI: 10.1109/TIM.2024.3425485

The sky seeing, which describes the Earth's atmosphere fluctuations, remarkably affects the quality of astronomical observations acquired by telescopes. Astronomers are inclined to measure the seeing using a ground-based tool in site selection studies to initiate an astronomical observatory. An all-sky camera can provide a powerful tool to measure the seeing conditions because of its low cost, lightweight, and easy deployment. However, the principal drawback is the low quality of the images, correspondingly complex to measure star characteristics which reflect the sky seeing effects. In the literature, several approaches are introduced for detecting stars; however, they are limited to high-resolution images, where expensive devices are used to collect the data besides the observatory domes, reducing the impact of light pollution. In this study, we propose two models based on machine learning, specifically fuzzy logic, to enhance the contrast of the all-sky camera images, precisely measuring the star characteristics that describe the seeing condition. The underlying hypothesis is grounded in the effectiveness of fuzzy logic in handling uncertain and ambiguous data. The adaptability of fuzzy logic to dynamic and unpredictable conditions, such as alterations in illumination and weather, makes it well-suited for improving the contrast of images obtained from all-sky cameras in various environments. We develop a nonparametric single-input-single-output fuzzy system with three linguistic values. Based on crisp, precise rules and membership functions (MFs), the fuzzy system does not have a fixed set of parameters and adapts its structure without relying on predefined values. We apply the two models to four datasets acquired from different sites located on four different mountains. We compare the models with benchmarks proving superior to our proposed models. In addition, we estimate the night sky brightness (NSB) to analyze the models' results using astronomical studies that show the concurrent results.