Clear and stable skies are essential for astronomical observations, directly affecting available observing time and data quality. Researchers from the Xinjiang Astronomical Observatory of the Chinese Academy of Sciences (CAS) have now developed an automated image classification model that can rapidly and objectively assess cloud conditions using images from all-sky cameras, offering a more efficient solution for continuous site monitoring.

The study, led by master's student WANG Siqi under the supervision of Prof. Ali Esamdin, was carried out on the Muztagh-ata observing site. The results were published in Research in Astronomy and Astrophysics.

The study resulted in the development of the automatic cloud image classification model, named ASCNet, which is designed to automatically classify all-sky camera images and maintains good stability even under complex illumination conditions.

ASCNet employs a complementary dual-channel feature extraction framework. In this design, ResNet captures global semantic information of the sky, and the ASCModule extracts local luminance texture features related to cloud structures. Combining these two types of information enables the model to discriminate between various cloud conditions.

In testing, ASCNet demonstrated strong consistency and stability in cloud image classification tasks. It achieved a consistency rate of approximately 92.7% compared to manual classification and effectively identified multiple typical cloud conditions. This indicates its strong potential for practical applications.

Automatic classification of all-sky camera images reduces manual workload and improves the efficiency of site monitoring. As astronomical observations become more intelligent and precise, this approach, which enables machines to understand sky conditions, is expected to play an increasingly important role in evaluating astronomical sites and supporting observational operations.

This work was supported by the "Light of West China" Program of CAS and the National Natural Science Foundation of China.