3. Can't be a joystick interface, since sometimes looking a specific spot on the field requires the operator to move to that spot by memory. Doing this with a joystick would be cumbersome, and look mechanical on the video.
I don't know how hard it would be to get image recognition / blob detection accurate enough to track a baseball reliably, given the high speeds and accelerations involved. One thought is that if you lose track of the ball or the angular rate is too high for your servos to track, you could zoom out to slow things down. Downside is that the ball becomes smaller and harder to spot, of course.
What about an analog joystick? A trackball? A mouse?
You don't say how big your cam is so no way to guess what it will take to move it. Servo City makes some large servo based cam mounts. There are large powerful servos available that could move most cams. That being said, have you done any simple testing to see if what you are trying to do is reasonable? The golf tourniments have the high end equipment to follow balls in flight and make it seem simple. One simple test would be to try to track the action only by using the cam video without any other visual references (cover your head with a towel so you can only see the video display). Some time back I made a joystick setup to control my servo based pan/tilt web cam. The cam moved quickly, but tracking anything moving fast just looking at the video from the cam is more difficult than it would seem.
I can't even conceive how to do blob tracking, it may beyond my current skill level.
QuoteI can't even conceive how to do blob tracking, it may beyond my current skill level.The best approach is to not do it on your own. Instead rely on solutions that already work, e.g. http://en.wikipedia.org/wiki/OpenCV.