Image recognition of star formations from a telescope using a star database?

Hi there, I am looking to use some AI tech on a stream of video coming from a telescope and wondering if someone could point me in the right direction?
There are apps, that can use smart phones and overlay stars you can expect to see at any viewing angle and time of day. These use star databases, along with the attitude from the smart phone’s sensors.
However, does any one know if such a thing has been done by an nvidia unit comparing a star database with a real time telescope image?
I had thought about superimposing a star map over a telescope image, using vectoring data from a multi axis sensor and comparing images but this would lead to clutter and confusion.
It would be far better if a low cost but powerful AI processor could give suggestions as to what it believes it is seeing in real time.
Scope: Whilst there are billions of stars and databases are massive, the AI would only have to pattern match angles between a few hundred stars in the region of space that the angle sensors are suggesting on the telescope.
In my short research, I have come across some very deep and math heavy papers that were written for guidance of orbital satellites and very big budgets. Nothing so far, on a low cost ‘point and learn’ telescope.