About calibration

Post Reply
jhjeon82
Posts: 24
Joined: Mon Jan 28, 2008 3:51 pm

About calibration

Post by jhjeon82 » Mon Jan 28, 2008 4:20 pm

Hi,
I'm new to this forum, and using twelve V100 cameras.

Once calibrated properly, the performance is pretty good.
But I have some questions and suggestions about calibration tool itself.

1.
Cameras seem to be recognized as red markers. By blocking those areas, I lose some pixels to track real markers. And they prevent us from calibrating accurately because sometimes some of them blink though they are blocked. This make "multiple markers" warnings. Why are cameras recognized as red markers?

2.
On the calibration video online, there were no limits for calibration data gathering. He moves wand until he thinks it's enough. But in the recent version, calibration process stops after it gathers some amount of data, probably 150 frames. It makes calibration very challenging because most data would not survive in the validation and there remain not enough points finally. With 6 or 8 cameras, calibration could be successful mostly. But with 12 cameras, I failed mostly. Belows are messages when the calculation process fails.

Only xx points survived validation and packing: probably not enough points for reliable selfcalibration
Matrix Error
No convergence in 30 iterations

Moreover, there are many cases that calculation proceeds and fails in another global iteration. The most miserable case is that calculation process succeeds and more than 3 points are detected when I put the calibration square on the ground.

I think these problems could be solved if you allow us to gather as much data as we want, like in the calibration video online.


Thank you, and I hope especially #2(it bothers me) could be solved properly.

johnou
Posts: 21
Joined: Wed Nov 28, 2007 6:09 am
Location: Australia.Sydney

Re: About calibration

Post by johnou » Mon Jan 28, 2008 4:55 pm

A)What kind of conditions are you capturing in? Is there any natural light in the capture space? That's include Moon Lights too.


B)If you are in the closed space with no natural light, what's ur machine spec?

Cheers

Birch
Posts: 1139
Joined: Thu Jan 30, 2003 5:00 am
Location: Corvallis, Oregon

Re: About calibration

Post by Birch » Mon Jan 28, 2008 5:17 pm

The default "Fast" calibration should normally produce enough points for the calibration to solve easily with 12 cameras. I would recommend trying #1 listed below along with John's comment #A above, and then trying #2 below if you are still not getting enough points.

1) If cameras are seeing the IR LEDs of other cameras, try increasing the blocking size to 15 or 20 before performing the "block all visible" operation. When an object becomes partially occluded its center of mass may shift slightly, sometimes the center of mass will move outside of the block region if the region is not large enough.

2) Have you tried selecting "Medium" or "Slow" from the Wand data drop-down menu of the Wanding page? It will allow you to capture more points. After doing that, you can also increase the number of points to 500 or 1000 on the Calculation page.

3) Also try to avoid blocking the view of the marker from any cameras with your body during calibration if you are not already watching out for that.

VincentG
Posts: 7728
Joined: Mon Jul 17, 2006 5:00 am
Location: Corvallis, Oregon

Re: About calibration

Post by VincentG » Mon Jan 28, 2008 6:15 pm

How do you have your cameras setup?

Setup volume, camera height?

jhjeon82
Posts: 24
Joined: Mon Jan 28, 2008 3:51 pm

Re: About calibration

Post by jhjeon82 » Tue Jan 29, 2008 10:20 am

In the closed space, I tried both cases, with light and without any lights.

I used 12 V100 cameras.

jhjeon82
Posts: 24
Joined: Mon Jan 28, 2008 3:51 pm

Re: About calibration

Post by jhjeon82 » Tue Jan 29, 2008 10:38 am

1) How can I increase the blocking size? Is there an option for it?

2) I always try with "slow, most accurate" option. Once calibration starts, progress bar starts to proceed. Can I increase the number of points manually? Estimating from the calculation page, only 150 points seem to be captured. (Camera #x, Width, Height, Frame Count : "150") I didn't count them, so it can be more than 150. Anyway in the calculation process, usually very small number of points survive. It's the reason that I want to capture as many points as possible.

3) I'm always trying to not block the view of the marker.

But, is the calibration working only in the case all cameras can see the marker? Wand data seem to be gathered if some of cameras can see the marker. And in the long run, I want to include obstacles like column in the capture volume. Is it impossible?


Thanks,

jhjeon82
Posts: 24
Joined: Mon Jan 28, 2008 3:51 pm

Re: About calibration

Post by jhjeon82 » Tue Jan 29, 2008 10:41 am

12 cameras, 6 tripods
Cameras setup was similar to the recommended 8 cameras setup on the webpage.
Setup volume is not big. 1m radius?

VincentG
Posts: 7728
Joined: Mon Jul 17, 2006 5:00 am
Location: Corvallis, Oregon

Re: About calibration

Post by VincentG » Tue Jan 29, 2008 11:38 am

With the cameras setup. Do you have them pointed to a center point, within the tracking volume, at about waist height?

Jim
NaturalPoint Employee
NaturalPoint Employee
Posts: 1395
Joined: Mon Oct 14, 2002 5:00 am
Location: Corvallis, Oregon
Contact:

Re: About calibration

Post by Jim » Tue Jan 29, 2008 1:37 pm

Hello:

The space between all the cameras is only about 1m, or your cameras are actually further apart and you are trying to get a 1m capture volume? With 12 cameras you can have a 2M capture volume for sure.

Calibration works better if you have cameras overlapped in their field of view. All cameras do not need to see the wand marker at the same time, but the more the better. If you set the calibration to Slow, it will collect more frames, but you also need to set the Calculated Point to be higher, like 500 or 1000.

Block setting is in the Calibration Setup Wizard, it is defaulted to 10x10 (I think) and 15x15 should work better for you, or even up to 20x20, just to be sure for this test.

Post Reply