Advice on Rigid Body design

[henderso]

Hi,

We are trying to track this homegrown head worn display (HWD):

http://farm4.static.flickr.com/3471/336 ... e0b3_b.jpg

(minus the Intersense tracker)

While we have obtained excellent tracking in our lab, we've had a hard time getting acceptable tracking in the actual task domain (military vehicle/cockpit).

We've calibrated it 100's of times (see setup below) and believe we have the best possible calibration under the circumstances. So now we are trying to make the rigid body as good as possible.

Any advice or pictures anyone has on tracking the head with an Optitrack would be very appreciated. Specifically:

- How many markers should we endeavor to have on this HWD?

- I understand we want asymmetric triangles...but how many? -

- Should the triangles be oriented in any particular direction?

-Are we going for a good cross section across each camera's FOV? So this would lead to multiple triangles, facing in orthogonal directions?

-If we use multiple triangles, do these have to be different in scale and leg length? That is can we have a 3cm,4cm,5cm triangle and a 6cm, 8cm, 10cm triangle and be OK? Or once we use a particular combination of leg lengths, do we need to find another asymmetrical combination?

Setup:

Domain: Left seat of turret (left seat) of LAV25
http://data.primeportal.net/apc/arrin_h ... 0hatch.jpg (not our photo, but representative)

Cameras: 8 x VT100 (We actually have 10, but found removing cameras actually helped -- possible by removing IR light that causes mean reflection on all the metal surfaces)

Average distance of camera from head: about 20 in

Camera arrangement: It a tough environment--can't place them exactly where we want to. Basically we have:

2 x cameras on top of vehicle looking down through hatch at head

1x camera in front of head at eye level along line of site

4x cameras off to the right, covering a 30 degree arc

1 camera off to right, perpendicular to user, at waist level

We don't have any cameras off to the left as the vehicle's skin is right off the left shoulder and I'm assuming within the operating envelope.

At them moment, we are using passive markers. We are considering trying active markers and disabling camera illumination.

Thanks

Steve

[Jim]

Steve:

Sorry for the late reply, this post was forwarded to me for follow up.

What a cool project! I really want to help to get this working. Optical tracking in the small volume is a bit on the tough side, but we can probably make it work, especially if you had it working well in the lab.

For RBs, what software are you using, did Doug get you a copy of the new Tracking Tools to try out? It is the sucessor to the RB software. Email him at support @ naturalpoint.com.

The reason I ask, is that the current software really likes to only have 3 and maybe 4 markers per rigid body and the new software is okay with 4-5 and maybe 6, which can help, it can also loose a few markers and still work well.

For this volume, I would have very few cameras, maybe 4? You will want to use small reflective markers, maybe only stickers? Turn down the Exposure to a low value, maybe 10? It won't take much IR to track at only 20". If you do not need 360 degree tracking, I would focus on creating only one RB and setting up the 4 cameras to track this RB in a volume, like on the side of the head with the most room, or from on top. You do want to spread the cameras out, they are more robust that way, the rays have more angular spread and the calibration will be superior. You do want about 60% or more camera FOV overlap, for calibration.

If you get it calibrated and tracking, kind of working, you can send us the data files and we can provide a lot more feedback. Try the new software first, then send us data files from it.

Jim

[henderso]

Jim,

Thank you very much. I will try it with the new software and then send you the data.

Steve