A cloudiness estimator for allsky camera images. The project was originally started for Petnica Meteor Group that sought to automate the manual process of estimating cloudiness in allsky camera night images for the purposes of calculating Zenithal Hourly Rate.
Given a timeframe, field of view of the sky, and allsky camera images, the program will print a table with cloudiness percentages as it changes over the time.
Detection uses a neural network trained to recognize clouds, including non-obvious ones in the night sky, which are detected by the lack of stars where the clouds are.
To see the program in action, run:
python cloudiness_estimation.py
Select internals/sample_images folder as the image source and set a wide date interval so it covers all the images.
Credits:
Project - Vladimir Nikolić (https://github.com/schutzekatze)
Neural network design - Leander Schröder (https://github.com/robinleander)
UI - Aleksa Tešić (https://github.com/Imafikus)