Development of a sky imager for cloud cover assessment

Abstract

Based on a CCD camera, we have developed an in-house sky imager system for the purpose of cloud cover estimation and characterization. The system captures a multispectral image every 5 min, and the analysis is done with a method based on an optimized neural network classification procedure and a genetic algorithm. The method discriminates between clear sky and two cloud classes: opaque and thin clouds. It also divides the image into sectors and finds the percentage of clouds in those different regions. We have validated the classification algorithm on two levels: image level, using the cloud observations included in the METAR register performed at the closest meteorological station, and pixel level, determining whether the final classification is correct.