Hi all,
our tracking setup encompasses 12 cameras which are adjusted around a table. We want to track the area above the table (hand interaction) as well as the heads of the users standing around the table. Therefore we have 4 cameras adjusted to the head area and 8 cameras to the table area. These two areas are slightly overlapping. By calibrating the setup with the 3-marker-wand we only achieve 'great' performance after a very long time of 20 minutes. Our last setup consisting of 6 cameras was much faster and accurate. We know that the complexity rises with more cameras, but maybe anyone has a hint, how to improve the calibration (time and quality).
Regards
matt
Calibration Time & Quality
-
Seth Steiling
- Posts: 1365
- Joined: Fri Jun 27, 2008 11:29 am
- Location: Corvallis, Oregon
Re: Calibration Time & Quality
Hi Matt,
Unless you require the head and hand capture areas to be seamless, you can actually turn each set of cameras into a unique camera group and then calibrate them separately. It's likely that having minimal overlap between the two groups could cause issues, if they are part of one large calibration.
Also, when you calibrate, try setting the desired quality to Medium first. Then, when it's achieved excellent or exceptional results, move it up to High. Then repeat with Very High. Stepping the quality setting up like this can help the calibration converge in trickier camera arrangements.
Unless you require the head and hand capture areas to be seamless, you can actually turn each set of cameras into a unique camera group and then calibrate them separately. It's likely that having minimal overlap between the two groups could cause issues, if they are part of one large calibration.
Also, when you calibrate, try setting the desired quality to Medium first. Then, when it's achieved excellent or exceptional results, move it up to High. Then repeat with Very High. Stepping the quality setting up like this can help the calibration converge in trickier camera arrangements.
Marketing Manager
TrackIR | OptiTrack
TrackIR | OptiTrack
Re: Calibration Time & Quality
Hi Seth,
thanks for your reply. We are not very familiar with camera groups, so there is one elementary question: Can these two groups be part of a common coordinate system, if the ground plane is only visible for one camera group?
I will check this stepping out - thanks!
thanks for your reply. We are not very familiar with camera groups, so there is one elementary question: Can these two groups be part of a common coordinate system, if the ground plane is only visible for one camera group?
I will check this stepping out - thanks!
Re: Calibration Time & Quality
You'll need to have the ground plane somewhere where both volumes can see it, or translate the volumes accordingly. They will contribute points to the same coordinate system.
However, you're upping the complexity of the problem by doing this.
The reason the calibration is taking much longer is because it's important that there is lot of overlap in the camera's views. When you have less, the data correlation between the cameras goes down and the convergence properties of the calibration process declines sharply. Ideally, you'll want to have lot of overlap between cameras, I would recommend adjusting the cameras so they are all looking more or less into a central area in order to make the calibration go smoothly. Hopefully you can do this while meeting your other design constraints.
However, you're upping the complexity of the problem by doing this.
The reason the calibration is taking much longer is because it's important that there is lot of overlap in the camera's views. When you have less, the data correlation between the cameras goes down and the convergence properties of the calibration process declines sharply. Ideally, you'll want to have lot of overlap between cameras, I would recommend adjusting the cameras so they are all looking more or less into a central area in order to make the calibration go smoothly. Hopefully you can do this while meeting your other design constraints.