Let me expand on that a little bit, though, so I'm not misunderstood: use the cam for image recognition, sure. But instead of piping the cam image to the projector, simply use it to map the augmented reality elements (visual tags like addresses, direction-path, and so forth) directly to the projector so they line up correctly with what you actually see through the lens. Add a digital zoom pip for when you want some far-off detail up close, I suppose...