s250e camera limits

[vvortex3]

Happy New Year everyone!

During the holiday break I happened to do some testing of my 3 s250e camera system to see what the limits would be. I am only using my system for head and gun tracking at the moment so the requirements, as far as I can tell, are more lax than what is required for full body mocap.

I tested the cameras in a triangle configuration that would have resulted in an 18.5 x 18.5 foot capture area using 7/16" hard reflective markers for everything. The result was that the Tracking Tools software was unable to calibrate as a result of wanding regardless of the camera threshhold or exposure settings. I would conclude that this size volume is not possible with the given hardware and markers.

I tried a slightly smaller volume at 13 x 15.5 feet and actually was impressed that the software was able to calibrate the volume (3 marker calibration). But the tracking was too inaccurate to be usable.

I am planning to try again at 12x12 and then 10x10 to test the real limits of a 3 camera s250e system.

I would also like to say that having 4 cameras would be much better than 3 (obviously) but particularly because you are no longer trying to fit a square capture volume into a triangle of cameras.

On another note, I really love the system and It works very well. I am satisfied with this system for Head and hand tracking purposes for real-time virtual reality applications. Keep up the good work Naturalpoint!

(videos at www.minus-reality.com)

[VincentG]

Dewayne,

Is the number you have listed the actual distance between the cameras, or is the 18.5 *18.5 capture areawithin the actual setup are of the cameras (closer to a 24*24*24 triangle)?

[vvortex3]

The 18.5 x 18.5 foot was the actual capture area that was attempted. I'd like to give this a try again with 1" markers. Can I obtain 1" markers that would work with the 3 marker wand and ground plane?

[VincentG]

You can get the 1" markers, on the website -

http://www.naturalpoint.com/optitrack/s ... cking.html

How big was the setup area?

You are using the default lenses?

[vvortex3]

I will have to do the calculations again to check what the setup area was for 18x18.

Tonight we just finished successfully getting a 12x12' volume to work acceptably with only 3 cameras and the same 7/16" reflective markers. It would seem that this is near the limit with this size markers.

We also rigged up some infrared LEDs to the calibration wand and were able to see it tracked as 3 distinct points at 75 feet. This is making me wonder what would happen if I build several bulbs out of a series of connected wide-angle LEDs and attach them to the 3 marker calibration wand and ground plane what size volume I could produce. I think I will be attempting this next and plan to order the parts tomorrow.

[vvortex3]

By the way, yes I was using the default 56 degree lenses so you could calculate the setup area based on that. One edge of the tracking volume was also the edge of the triangular camera configuration (resulting in relatively inaccurate tracking near that edge).

Something like this:

Code: Select all

    
    /  \
   /    \
  /______\
 / |    | \
/__|____|__\

[vvortex3]

From what I have read so far, it seems that a trackable volume must be calibrated within the area of overlap between all cameras involved with that particular volume.

That said, it seems that markers can still be tracked outside of this volume as long as at least 2 cameras can see the marker. Is this true?

Are there limitations in the software that would prevent me from calibrating a volume that is much larger than the stated range of the cameras? After proving that a single camera can see the 3 marker calibration wand at 75 feet as 3 distinct points (replacing the markers with LEDs) I am going to attempt a very large volume using active infrared LEDs.

I am planning to use epoxy putty to place the wide-angle LEDs (sold by NaturalPoint) in a ball-like configuration so they will point in as many directions as possible. What do you think?

The rigid bodies that I am tracking will consist of 3 markers each in a 1x1 foot configuration like this:

Code: Select all

         O (4x infrared LEDs)
         |
         |
O--------+-----O (4x infrared LEDs)
(4x infrared LEDs)

This is the rigid body that is attached to the head mounted display.

[beckdo]

Hey Dewayne. This some nice work you're doing here. I wanted to give you a couple pieces of information so that you don't have to second guess what is going on.

First off, there are no expectations placed on the wanding data or solution. What I mean is that the algorithm will take whatever data you give it and try to converge on a solution. Wanding 75 feet from the camera will not prevent a solution if the data was good enough.

I warn you though, 75 feet is not going to work unless you fashion a much larger wand. Otherwise the points will be too close together to yield enough meaningful data to converge on an accurate solution. Additionally, you'll have to relax most of the constraints in the Wand Acquisition section.

The process calculates an initial solution and then through multiple numerical methods refines the solution. So if you've got the camera arrangement such that there is only a minimum amount of overlap in the cameras views there might not be enough wanding information to yield quality results.

You only need a minimum of 2-camera visibility anywhere you want to track. However it's important to have sufficient overlap in your camera placement. With more overlap the wanding data will exhibit stronger convergence properties. This is not something you can observe with the naked eye but does exist mathematically.

[vvortex3]

Interesting. I thought the 3 marker calibration wand had some sort of hard constraints preventing me from fabricating a larger one. If I fabricate one that is say twice as wide, are there any things that I have to do for it to be accepted as a calibration wand? Do I need to maintain the scale of the marker placement? What effect does a larger wand have on the resulting trackable volume?

What about the ground plane?

[Jim]

Hello:

Doug is out today, so I will jump in.

You can set the wand size in the Calibration settings in TT. Yes, the ratio between the markers must remain the same.

The Ground Plane can also be changed, ratio the same, and entered into the settings panel.

You will also need to change the Point Cloud settings so that it will not reject tracked rays that are over 10M, because it will toss them now.

I have made IR LED clusters out of our epoxy domed LEDs and they do work quite well.

I would also suggest just trying the 1" markers out, the range will increase significantly with normal passive tracking.

Have fun!