Please see a short explanatory video on YouTube.

Additionally, this tutorial is a really useful beginner's guide to OpenCV classifier training.

Contents so far:

• Stellarium scripts used to capture thousands of images from Stellarium in order to be processed into negative image datasets for machine learning training.
• Zipped folders containing negative image datasets, as well as bg.txt files, and python programs used to create these.
• Python programs used to create the positive images used for cascade training.
• Image files of the fiducial markers applied to starfields, to identify the patterns of bright stars that the machine learning relies upon for the identification.
• A sample set of 31 trained cascades for the northern celestial hemisphere.
• Python programs used to test the trained cascades against a supplied starfield image.

What next?:

19/08/19, I have finished working on this project as part of my University course. I hope to be able to spend further time on it as a hobby in order to keep developing the system, there are lots of improvements and additions I would like to have time to make. I hope that this repository may be of use to someone, and if you have questions please contact me, I will continue to monitor and work on this project. The best source of reference here is my MSc Thesis itself, which can be found above.

25/05/20, I've been putting more thought into the potential improvement and applications of this project. I hope that the Instructables writeup will help other people find this repo, and hopefully we can work together to develop this further!