Tracking Camera

I have a need for an automatic tracking camera mechanism. I want to use a
video camera on a tripod which will pan (tilt is not needed) in the
horizontal plane over about 30 - 40 degrees. The camera is to automatically
track a couple in ballroom dance competitions when there are between 6 to 10
other couples on the floor. I invisage some sort of transmitter unit (small)
which is worn by the couple, which is tracked by the camera mechanism which
will follow the couple around the floor. My partner and I are competing and
we are having trouble finding someone to operate the camera for us, as most
people we know are on the floor at the same time as us. I am hoping that
someone in this group would offer a few suggestions. Having built a number
of robotics projects, I'm competent enough to put the electronics &
mechanicals together but I'm lacking in the ideas and design area on this
Murray McKenzie
New Zealand
A person who works with his hands is a labourer.
A person who works with his hands and his brain is a craftsman.
A person who works with his hands, his brain and his heart is an
Reply to
Imagineering Unlimited
Loading thread data ...
Well, it is certainly a novel requirement ! The 'best' answer is probably to cut back on the ballroom dancing, so as to give you more time to cultivate friends who you can train to operate the camera ....
The technical solutions are not easy. Some sort of beacon (IR) is probably the simplest, but unless you mount it on a mast (mind boggles), it will likely be obscured from time to time. Another route is one of the indoor 'gps' type systems mounted on the couple, transmitting a position to the camera, which is then aimed 'blind'. If you wear a distinctive colour, then there are video systems that can extract specific coloured regions from an image. There are systems used for cctv surveillance that are claimed to be able to track people across images and multiple cameras. Then there are those systems used on police helicopters that track cars or people.
None of this is remotely easy
Reply to
Dave Garnett
You could set up a camera with pan/tilt/zoom controlled over the internet, assuming you can arrange internet access at the venue. Then your potential pool of camera operators increases enormously - maybe you have dancing friends in other geographic areas who won't be in your competition. You could offer to pay someone somewhere if there are no volunteers.
Mitch Berkson
Reply to
Mitch Berkson
The tracking of a specific colour is actually quite simple with the right technique. Once I realised that using HLS colour system rather than RGB was a better idea, it was simplicity itself to do. I am using it on a robot prototype I am in the process of programming, and it works extremely well.
The only issue is one of speed. At the moment the prototype is in visual basic (no laughs please!) and it can manage about 70fps on a 320x240 capture image, which is actually just under 4 times faster than the digitiser can supply the frames, but this is on a 1.4ghz machine. With something like a mini-itx board (you cant really take a full ATX desktop to these events I guess) you wont get near that. When I get some time I will port it to C++ or something for use in linux which should help.
My only other suggestion would be perhaps to have some sort of directional radio antenae aranged in a small radius semi-circle then just rotate the camera to where the most antenae are getting the strongest signal. With some sort of software that could "predict" the movement of a dancing couple it might be workable (for example dancers dont generally move 20 degrees across the frame in a short time frame) even with a signal that gets blocked a lot.
Reply to
Matt Dibb
Thanks for the ideas Guys, Unfortunately we all wear black or midnight blue. I didn't think it would be easy, I think that severe bribery is the only way out of this dilemma.
Reply to
Imagineering Unlimited
We did something like this before. What you could do is seperate the tracking from the 'filming'. We used a camera above the floor and labeled the couple with a small IR retro reflecting label. Then float the floor with IR and use a IR filter before your (B&W) tracking camera. Looking from above makes tracking much easier. With this information you can stear the other camera.
Good luck,
Peter KITT Engineering
Reply to
Peter van der Vos
Just had a non-roboics thought. Set up two cameras, each viewing part of the dance floor so between the two you get reasonable coverage. Perhaps a slight variation is to put each camera relatively close to the dance floor, but with a wide enough angle of view to see the entire floor. This way you will get some close-ups as well as the wide shots.
For the robotics approach, even if you can scan for a particular color, it might be viable to use an infrared LED light source (in your dance-partner's heair dressing?). If I recall correctly, I think one of Sony's cameras has a "night vision" setting where it removes the infrared filter normally in the light path so that the camera becomes sensitive to infrared.
Here are some systems that might have some possible use... but I have my doubts any would work well. I can't say I am very familiar with any of them anyway, so perhaps someone else will get inspired with them.
-Logitech now has a Quickcam that can follow a person.
-There are also a low-cost program cam2pan program ($20), which will track a colored dot in an image. It is used to control a mouse by tilting your head.
-Carnegie Mellon University has their CMUcam Vision system available for purchase.
formatting link
Reply to
Have a look at the cmu cam for a few ideas. It can track colors
Reply to
Alex Gibson
But an IR beacon may show up as an annoying bright spot in your, and every other camera without adequate IR filters. Simple to test with a TV remote control...
Chris Stratton real address is my last name at mit
Reply to
Chris Stratton

PolyTech Forum website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.