Motion Tracking

[JEC]

I've been asked to design a system which will reliably track a person walking back and forth across a room, and move a video projector back and forth so that the center of the projected image is (more or less) located on the center of the person's body.

It's for a Modern Dance installation.

Picture a video projector mounted on the head of a standard photographic tripod and you'll have a good idea of what I have in mind.

The projector will have to pan back and forth over a 100 degree arc, but won't need to tilt up or down. Control will be via geared steppers & position sensors, or plain servomotors. And trapezoidal accelleration curves and all that.

I've considered having the performer wear a bracelet or belt containing several high-powered IR LEDs. I could then mount a generic video camera on top of the projector so that they move together. An IR filter on the front of the camera would let me only 'see' the bright spot.

If I can capture the video signal 5-10x per second and break it down to it's basic pixel elements, I can then send pan the projector back and forth until the captured hot spot matches the 'x = halfway' position in the pixel matrix.

Thus, the center of the projected image will more or less match the X position of the dancer's belt/bracelet.

What am I missing? Is there an easier way to do this?

Thoughts?

John

[Dwayne]

What am I missing? Is there an easier way to do this?

How about heat sensors? If mounted on a rotatable hinge, set off my about 30 degrees. Kinda like the following:

___
/ \

Object.


measuring each of the devices, one will probably show more "heat" than another, and turn the camera to that direction. When they are equal , they are in the middle.

I have never done something like this before... Maybe someone has, and can say "Dwayne aint going to work" or come up with a much better idea than mine.

I think you can get infrared heat sensors at cheap store for next to nothing.

Dwayne

[JEC]

I've thought about heat sensors, actually.

The faculty I'm working with on this wants a bit more precision than just 'point it at the warm blob over there.' How much precision do you think I can get with those devices?

Ultrasound would work if I only need to pan back and forth. I pulled some nice papers yesterday that talk about localizing a source based on a plane wave arriving at two closely spaced detectors. Depending on the location of the source, the detectors will receive the same signal either in phase, out of phase, or something in the middle.

So the source of a train of pings would be fairly easy to calculate.

If I mount the whole mess fairly close to the ground plane, I can avoid a lot of ambient reflections - similar to how a PZM microphone works.

Unfortunately, it looks like she may want the projector to tilt as well. Which either means moving to a video based system where I can look for a hot spot, or add a second ultrasound sensor pair mounted perpendicular to the first so I can calculate the angle above and below a plane.

....X
.
.X....X
.
....X

JEC

[Archilochus]

It would be a whole mess of wiring - but how about a grid of pressure sensitive strips (used for industrial control/automation) - or maybe a grid of active IR 'beams'.

Arch

[Dwayne]

Hello JEC,

JEC>>How much precision do you think I can get with those devices?<<

I don't know... But I would venture to guess you can become fairly accurate with them. I have a friend that has experience with them... I will ask.

Dwayne

[mister_e]

Unfortunately GPS are not so accurate as you want .... i'm joking

for sure a load of infraled emmiter and receiver in an matrix config will do the job. Acting like touchsceen. but what about the dimension of the sceen ???