In February, while on winter holiday trip, I visited aurora researchers at The University Centre in Svalbard (UNIS). We talked about machine learning and stuff. They recently set up a colour camera at Kjell Henriksen Observatory, and are interested in detecting and recognising auroras from images in real time.
While on a plane I tried a few quick solutions based on pixel thresholding and pixel clustering, they seem to be working reasonably well.
Here is an example of a threshold based detection:
For those, who ar interested to try, I put the code on GitHub.