Yeah as soon as I saw the title I made the assumption that it'd be automated.

I can't imagine doing this manually - though it'd take me more time to write a Machine Vision solution rather than just doing it manually. :P