The use of sky images in solar radiation intensity estimation has been one of the most studied topics in the literature since it improves the estimation results. The first step in processing sky images with image processing methods is to separate the pixels in the images as clouds or sky. This process is known as cloud segmentation in the literature. In this study, the sky is photographed using the sky imaging system installed at Afyon Kocatepe University Solar and Wind Energy Application and Research Center at times with different clouding characteristics and cloudiness rates in Afyonkarahisar Region. The photographs are divided into 25 parts, and small sky patterns are obtained. The pixels in the obtained sky patterns are manually segmented, and a cloud segmentation dataset is created for future studies. Since the resulting dataset contains high-resolution images and prelabeled data, it can be used to obtain more accurate results for the segmentation process and allows learning algorithms to learn faster. The dataset can be used by researchers in studies such as solar energy forecasting, meteorology, and weather forecasting, and the dataset in this paper will be shared with researchers upon request.
Read full abstract