You say it's not too hard, but I've actually worked on OCR projects, so I know how difficult it is to do this kind of pattern recognition. And it's not the programming--that's the easy part: it's coming up with techniques that actually work for recognizing stuff. Detecting lines is one thing: recognizing a flat sign in the middle of a bunch of staff lines with note stems nearby is horribly difficult.
Are you at all familiar with printed music? The symbols used vary all over the place; one publisher's sharp sign may look a lot different from another one's.
I'm sure it is do-able; just not sure how realistic this would be for a one-semester student project.