Clear Sky Detection with Deep Learning and an Inexpensive Infrared Camera for Robot Telescopes

Abstract

Automation methods can be applied in robot telescopes for autonomous operation. The clear sky detection is a key point for the telescopes to observe night sky in adequate conditions and close their domes in the case of a potential rainfall. The authors of this paper taken images of the night sky by a low-cost camera of a Raspberry Pi and trained a convolutional neural network (CNN) to detect clear sky on the pictures. The developed model reached 95.74% accuracy on unseen samples hence it was proved to be ready for production use.