Statistical Approach for Sky Clouds Density Classification

In the renewal energy mined from solar panels is essential to know the future amount of produced energy. The sun is a relatively stable source of energy, and we can precisely estimate extra-terrestrial sun intensity based on the hour of the day, day of the year, and respectful distance from the sun. The incident solar radiation hitting the Earth is affected by the Earth's atmosphere, climate, and the density of clouds. We need to predict sky clearness, respectively the density of clouds in the sky. This paper deals with sky clouds density estimation using a statistical approach. The data are acquired by a terrestrial fisheye camera facing the sky. In the first step, the various sky types were manually annotated to segment sky into artifacts - sun, clear sky, partial clouds, clouds, and terrestrial background. We used the set of Gaussian Mixture Models for the classification of such artifacts. We optimized the number of components in mixtures appropriate to different class requirements. The result of modelling should be the prediction of clouds density depending on the image captured by the fish-eye camera.