Also nothing to do with cell phones, but I did a motion tracking pan/tilt camera a few years back. Programming the interface with the camera was the hardest part. The motion tracking in itself was minimal. I was going to try for object recognition but ran out of time cause the Sony EV-30 (I think?) took forever to get at a reasonable price. It's a well documented camera though, and lots of people write software for it, so finding references was pretty easy.
Another plus, I sold it for $100 more than I bought it for on eBay
It was a fun project, and a hit at our presentation. Everyone walking around look at the different exhibits, and our camera was going apeshit tracking them as they walked by.
edit: the camera itself also has it's own motion tracking/object recognition, so you'd have something to gauge your own software against.