DPI-ACES Alfa. Digital Panchromatic Image - Automatic Cloudiness Evaluation System

We have developed an image-processing procedure to obtain an automatic cloud-coverage evaluation. Images to be analyzed are captured by a ground-based digital camera, equipped with fish-eye lenses. The images are in the visible-light spectrum; the cloud coverage evaluation is performed taking into account RGB and Hue parameters. By setting appropriate RGB values the algorithm provides for the discrimination of different colors, separating typical cloud palettes from those of the clear-sky. When a solar flare is present in the picture, it can be recognized and isolated by setting an appropriate Hue demarcation value. The algorithm has been implemented in a JAVA program, which was then tested against 502 images. The program output, expressed in octaves of coverage, was verified by comparison with cloud coverage evaluations made by three, independent human observers on the same images set. The automatic system accuracy proved higher than the human operators' one throughout the entire image series.