8Cam OptiTack+ARENA Test #3 (Jan.22.2008)

[johnou]

Before we start I will thank Optitrack for their support leading up to our 3rd capture.

Our aim for our 3rd shoot was to come up with the best possible setup and environment to allow the best possible capture volume.

- 10x10 metres controlled TV studio.

- The recommended 8 camera setup which is shown below.

- Each camera individualy focused on the same point.

- Machine used: Macbook

- Powersave mode on USB root hubs off

- Taken care with tension on USB connections to camera.

- Calibrated based on video demonstration.

Here are some photos of our setup:






Results:

Generally great quality of captures at first. Near perfect for half of day. Then the quality got slightly worse. For majority of time it appeared to be one camera -our licensing camera stopped capturing. We were certain that this was the only problem camera and out of the the rest, this one did fail to capture the most times. But then that camera fixed itself making way for another camera to fail. Forced to restart many times.

Taking on board the classic demonstration video on calibrating, we aimed to get the best possible calibration result to try and maybe fix this problem with our cameras. We calibrated - drawing crosses at each camera after starting low then moving up - trying to fill the screen with green. The results with this method was worse then what we used to! We normally get all excellent to great by staying low, without crosses at camera and not moving around as much. We made sure we didn't obscure the wand yet our results with the method in video after 4 attempts were varied with only a couple of excellent ranging down to fair overall.

At end of day, though our TV studio had no natural light source, we started to see 'noise points' appearing on our figure. From the 34 points to 39 and constantly jumping up to 42.

We also tend to think that anymore than 60 secs of capturing, leads to more errors - the figure jerks, and points get re-arranged - as if the data becomes too much or something.

Overall though, I feel that these problems are quite minor. The Optitrack system is a solid system. We are getting roughly the results we need which is excellent. It has blown me away with it's power and ease of use. But these problems make me aware that it is all just a process of motion capture - no shoot is always 100 percent perfect and cleaning is always needed.

We have tried so hard to replicate the same capturing quality as found in your video of the 'female performer'(Sigraph 07) on stage. If we were to upgrade to a 12 camera system, we would want to see better results and stability of our 8 camera setup. We are asking for the exact dimensions, height and width of the setup shown in the video as well as a camera settings eg. Exposure and threshold settings.

[color:#FF0000]We would need to know what governs the use of camera settings eg. What conditions would require the threshold and exposure at a particular setting. I guess we are looking for a sort of guideline to get the best possible capture using these setting with respect to the conditions on the day. [/color]

[VincentG]

A couple of quick thoughts......

1 - Looking at the marked out volume area, it appears like the overall setup is closer to 5m*5m, correct?

2 - Have you tried turning on unmarked rays, from within the software to determine where the stray points are coming from?

3 - Does the tracking appear to get better, if you do a calibration again?

4 - Could you send us some raw data to look at?

[johnou]

Here is your answer:

1. No. The marked area on the ground is 2.3 * 2.3 metres.

2. No. This will interesting to see next time.

3. No. We get different results when we change the settings of exposure and threshold. Do you have any explanation or guidance on what to change here? We keep it pretty standard - exposure 55, threshold 150.

4. Yes. We will try to send raw data to you when we have the chance.

Cheers

[Jim]

Hello:

Thank you for the excellent post and pictures! I think this is a very good point to work from to get even better results. We will compile the data from our I/ITSEC performance, I think that is the right one, and post it on our site and link here, the PT2 file will show you the camera positions. We will also post more data here on how the cameras were setup.

For the space you have, 12 cameras probably would help with your coverage. We have run the system for days, with no cameras failing, but I have also heard of one other customer that has had cameras quit randomly. We are looking into this. Finally, you might want to upgrade your capture computer, as with 1minute+ captures, you could need more RAM and processing power.

Also, the cables could be strain reliefed more to the tripods, no tension should be on them at all.

For the calibration technique, you are right, different things work for different people. In general, it is better to cover more of the volume, as it will be more consistent over the capture space, rather than just the area you wanded. For now, I would stick with what works for you.

Yes, using Untracked Rays, is very helpful to find sunlight, reflections, or just poor tracking.

We also have a new release of Arena scheduled for early next month. This should be a significant upgrade and allow for even better and more reliable tracking. There are even some features that will really make Arena a more mature product.

[johnou]

Thank you for all your support for now. We will continue to take your valued advice on-board for the next capture. We are very excited about what is to come for the Arena upgrade. We also very eager to find out more about your excellent setup for the I/ITSEC performance.

Our latest capture day used a Macbook which had 2 gig RAM. Our second capture session had similar problems over 60 seconds with 4 gig RAM. Must admit, the problem is quite minor (just a few jerks and so forth, nothing major or serious but does affect capturing)

If possible, as a suggestion, a overall percentage complete indicator on the calibration process would be very much appreciated in Arena

Cheers

[Jim]

Hello:

Thanks for the suggestion on the calibration status, I think that we don't really know when it will finish, as it is optimizing until it comes to the best solution it can, but I will run it by the guys.

One more idea to try, make sure you turn off all virus software or update checking software that can run in the background, they can cause the CPU to do other tasks when you aren't aware of it.

I hope that our new release will fix some of the capture problems when you are doing longer takes.

We will post the PT2 data from I/ITSEC soon.

[yoshi]

John,

If you are experiencing jerky motions only for very short moment around 60th sec, I recommend leaving "Record Length" checkbox on. This way some pointer is pre-allocated before the capture and have less interference with capturing data

[johnou]

Hi Yoshi,

That's what we have find out, we did limit 60s per capture.

[Jim]

Hello:

Here is the link to the files for the I/ITSEC video, all are included in one big zip file.

http://media.naturalpoint.com/video/opt ... EC2007.zip

[johnou]

Hello Jim

Thank you for the I/ITSEC files. They have been very helpful and we just have a few more question.

Camera settings - Threshold: 200. We notice when we mouse over, how it says 'Points must be brighter than this threshold in order to be visible.' We understand what this means but we want to know what are the effects as we go higher or lower. What conditions govern adjustment of the threshold? As in, do we lower the threshold in brightly lit studios etc?

Exposure: We notice that adjusting our exposure higher makes our points stand out more. We have observed that the I/ITSEC exposure is at 40. When we see the scale of you setup and the distance between the camera and the points, we would have expected the exposure to be at around 70 to 90 but your capture is fine. What are the kinds of things that govern the adjustment of exposure?

We just need a bit more clarification on these two settings. Thank you again for the files.

Cheers