A Cloud Cover Measurement Index Based on All-sky Imager Images

All-sky imager (ASI) images are an important means for the astronomy community to monitor cloud cover at present, but there is no definite cloud cover calculation index. Therefore, this paper proposes a new quantitative index for cloud amount measurement-Cloud Distribution Density of ASI images (ASICDD), and establishes an automatic classification system for ASI images based on this index. Firstly, all the images in the dataset from the Key Laboratory of Optical Astronomy at the National Astronomical Observatories of Chinese Academy of Sciences (CAS) are denoised and the cloud area is segmented from the sky by OTSU algorithm. Secondly, the cloud amount for ASI image with the background remove is calculated by ASICDD proposed in this paper. Finally, we use four traditional classifiers (SVM, KNN, Decision Tree and Random Forest) to classify ASI images automatically according to the calculated value and evaluate the performance of each classifier. The results show that ASICDD can be used as a numerical index for judging the cloud cover of ASI images; the automatic ASI images classification system based on ASICDD achieves a high recognition accuracy rate. Meanwhile Random Forest has the best classification effect-the recognition accuracy rate of various cloud images has reached more than 95%.