s250e camera limits

[vvortex3]

I have finished fabricating a calibration wand and ground plane that are twice the dimensions of the originals while maintaining the same scale. I will be attaching the LED clusters this week and most-likely testing it all out this coming weekend.

[vvortex3]

Where can I find some documentation regarding the settings in the "Wand acquisition" section? What settings would you reccomend if I am attempting say 40x40 or 50x50 feet and have doubled the calibration wand and ground plane size?

Also, what settings would you reccomend in the "Point Cloud" section?

[beckdo]

That's a great looking wand. What is the end to end distance as well as the distance to the center marker and I will put together a list of settings.

[vvortex3]

Hey Doug, All of the measurements for both the ground plane and calibration wand that I have fabricated are exactly double.

I have tested the system this week and determined that the maximum trackable volume I could produce that was reasonably usable was about 16 x 18 feet with 3 cameras. I was able to calibrate the system and even set the ground plane at even 25 x 25 feet but the tracking was unusable. Also at any more than about 18x18 feet the system was very tough to wand properly to the point that I could set the ground plane.

I think that about 16 x 18 feet is the maximum that 3 cameras could be stretched to. That said, I think that it was a matter of not having enough cameras to properly calibrate the system at such a distance rather than the camera's range that prevented me from getting 20x20 or even maybe 25x25 feet. I would very much like to try this again with a 4th camera as soon as I can afford to purchase one. Maybe I can get a special deal for doing all of this testing?

Here were the settings that I used to get a usable 16 x 18 foot trackable volume from 3 cameras. The setup area for the cameras was 33 x 50 feet:

calibration -> wand acquisition:
wand length: 800
center distance: 280
min object size: 1 pixel
min feature size: 1 pixel
min object roundness: 0

point cloud:
enable point cloud reconstruction: yes
maximum ray length: 25 meters
min marker size: 1
min roundness: 0
max x meters: 25 (I think these 3 settings were not actually used)
max y meters: 25
max z meters: 25

Camera settings:
camera illumination: 0
threshhold: 53
fps: 250
lights dimmed

I should note that the cameras were able to clearly see my calibration wand's 3 points in precision mode from as far as maybe 75 feet since it was made from LED clusters, but it could not accurately calibrate them from such a distance which I'm concluding is due to not having enough cameras.

[beckdo]

Hi Dewayne,

I apologize for the slow response. Your large sized wand looks good and I think it could give you reliable and repeatable results if you used it with passive markers.

Ideally we want a point source, but having a spherical marker is a fair approximation so long as we can accurately locate the marker's centroid.

The wand you're using with active LEDs does approximate a passive marker but unfortunately as you've likely found is that it's very hard to put the LEDs very close together and with an even light distribution when viewing the active marker from different angles. As the cameras see the LED cluster from different angles the uneven LED light and intensity variations will certainly impair the ability to find the centroid location throughout the sample set to an acceptable level of error. This is ultimately manifest itself in lower quality calibration results.

One final comment, I wouldn't decrease the minimum feature size below it's default. This setting is independent of the wand size and is effective is removing data that will likely make the calibration result worse than if the samples were rejected. Also, disabling the filtering of markers that are not round is not recommended. I'm assuming you needed to do this because of your active wand, which is symptomatic of why it's not working as well as you'd like.

Hope that helps

[vvortex3]

So I have ordered a 4th camera and am planning to try to set things up using the following layout. Any input/suggestions would be very helpful. The room is ~40' across and I'm planning to wall mount the cameras. The ceilings are 20' tall. All measurements are approximations.

[vvortex3]

It turns out that I now have access to install cameras on the roof itself. This seems to result in less distance between the camera and what is being tracked, but a greater chance for occlusions. In my case there should not be many occlusions from a top-down viewpoint. The room's dimensions are 40x40 with 20' ceilings. What is the maximum usable space with 4 cameras?

[vvortex3]

Just a quick update. Here is the top-down configuration using 4 cameras that I am planning to begin setting up today. Most of the math is approximations so please be gentle.

[NaturalPoint - Mike]

Hello Dewayne -

I've had a chance to look at your diagram, and compare it to the specs on the s250e. It appears that you should be fine with that placement, however you're right at the limit! I'd like to know how this works out for you.

[vvortex3]

Well we ended up doing the installation over the last week or so. The 4 cameras are spread 28x28 feet apart and are located ~17' off the ground. The resulting volume was actually 15x15x8 and not 20x20x8 as we calculated assuming the cameras were 20' off the ground (it turns out that the ceiling was not actually 20').

I am planning to write up a detailed article about the installation on my blog (www.minus-reality.com) within the next couple days. But, to summarize, the volume would not properly track with the standard 7/16" markers. Additionally, the volume would not calibrate properly when wanded with the calibration wand which has 7/16" markers attached. I ended up using 3/4" markers for EVERYTHING and also using my double-sized calibration wand with 3/4" reflective markers for calibration.

I even tried attaching 3/4" markers to the standard calibration wand, but the double-sized wand with larger markers resulted in a much better calibration than the standard sized wand. I think Naturalpoint should consider selling a double-sized wand for this reason specifically.

The volume works extremely well and about 30 people have tried the simulator so far with pretty good feedback. I will be posting videos in the next couple days.