Go Down

Topic: Remote controlled camera (Read 899 times) previous topic - next topic

CaptRR

Hello,

I am looking to make a remote controlled camera for the baseball games we cover.

Goals:
1.  Has to be able to move a standard camcorder, at a fairly decent speed to track a moving ball.
2.  Has to have a fast response, it would do no good for an operator to turn the camera to follow a player or ball only to have it respond a second later.
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.

Questions:

1.  I looked on the net and found some people that were using a wiimote and some servoes to point a webcam.  I was wondering if it would be better to emulate them and use a wiimote or something similar for the operator (could put the wiimote on a standard tripod mount to emulate the feel of a the camera), or would it be better to use encoder or even potentiometers?

2.  I don't think I have any servos that would move that much weight around easily, I do have a bunch of NEMA 250oz in stepper motors, and a controller, plus I could use microstepping so I don't have to gear them down.  Could I use this?  Or would it be better to find a more powerful servo?

I have a cnc machine so cutting out the case and some of the parts should be pretty easy, though I know those are some famous last words.



Udo Klein

Quote

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.


What about an analog joystick? A trackball? A mouse?
Check out my experiments http://blog.blinkenlight.net

PeterH

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.
I only provide help via the forum - please do not contact me for private consultancy.

CaptRR


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.


I can't even conceive how to do blob tracking, it may beyond my current skill level.  As it stands we already keep the camera zoomed out as default and only zoom into the action as it happens.  If I could figure out how to track a blob like that accurately I could do all sorts of cool things :)  (on screen overlays of where pitches are happening, average pitching speeds, heat maps, ect.)

Quote

What about an analog joystick? A trackball? A mouse?


I was thinking about this as well, and it might be a good idea, probably would have to go directly off the encoders of the mouse / trackball to get the reaction time we need however.  I think a computer would probably add too much lag to the entire system.  I wonder if I could tap into an old ps2 mouse?  Its definitely an idea, and their is no reason why I couldn't add some kind of wiimote to the system later if needed.

zoomkat

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.
Google forum search: Use Google Advanced Search and use Http://forum.arduino.cc/index in the "site or domain:" box.

CaptRR


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.


You bring up some good points.  I can say I do think its possible since I have had to run the camera myself a few times, and had to move it to follow the ball, while looking at a completely different screen (I also run the switcher, and stat program, and do color commentary).  Its another reason I didn't want to use a joystick, since following the ball is as much feel as movement.  I.e.  When you hear the crack of the bat (which we can hear with our audio equipment) automatically start moving the camera in the expected direction.  Using the camera zoomed out helps this, and usually we zoom in only once we got a clear indication where the ball will go.

You also make a great point about golf, but we have done golf, as well.  Golf balls are allot harder to track with a camera.  They are smaller, often further away, and worse of all once they get in the air, their is no frame of reference, no one to look to see where the ball is going.  In baseball, or football if worse comes to worse you can always zoom out, and look at where everyone is looking and point the camera their :)

The camera itself doesn't weigh much, only about a 1.5 - 2 lbs.

Those mounts you mentioned from servo world, interest me greatly.  In fact they look almost perfect, as long as the speed can keep up, again, were not talking about super speed here, and the further away things are the less speed the camera mount really needs.  In fact the only time you really need allot of speed is when the ball is passing very close to the camera or crossing it view at 90 degrees.

Udo Klein

Quote

I 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.
Check out my experiments http://blog.blinkenlight.net

CaptRR


Quote

I 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.


I have to admit it would definitely be high on the cool scale to lock onto a players helmet and have the camera follow him wherever he goes.  Still I I think I will take this once step at a time.  The the camera remote control working fist, the play around with adding cool things to it later.

Go Up