forums.naturalpoint.com

Hi folks:

I've looked through your FAQ and the SDK documentation, and didn't find any mention of this topic. Does your SDK automatically compensate for the intrinsic camera parameters (lens radial distortion, pixel array decentering, and skew)?

Thanks, Scruff

There is not any support for this kind of correction in the current version of the SDK. There has been considerable customer interest in them, so we are reviewing them as possible features in future versions.

Thanks Brich:

We have one member of the yahoo motion capture group ( http://movies.groups.yahoo.com/group/motioncapture/ ) using Dr. Tomas Svoboda's multi-camera calibration toolbox for their OptiTrack implementation. If someone has access to Matlab, his toolbox is very user friendly and determines both the intrinsic and extrinsic camera parameters. All you have to do is wave a laser pointer through the capture volume, and process the laser position data through the calibration toolbox.

Dr. Svoboda's web site is at:

http://cmp.felk.cvut.cz/~svoboda/SelfCal/index.html

Since the camera calibration toolbox uses the synchronized image frame paradigm for it's processing, the OptiTrack user was simulating synchronized marker frames from multiple cameras by interpolation (the same technique I'm using for my mocap rig).

Regards,

Michael Miles

Scruff and Brich,

Thanks for posting this. I'm also trying to calibrate my Flex 120 camera.

I'm using the Roger Tsai method, and need the intrinsic camera params. I'm assuming using 2.2 mm by 1.8 mm for the image sensor and 355x288 will work.

I'm confused, though, how the SDK works for vector tracking if the cameras have not been calibrated. Does it assume a "center" center point, no distortion, and the nominal 3.4 mm focal length?

Jim

Correct, lens distortion is not compensated for in the Vector tracking when using a C120 camera. It is corrected for TrackIR 4 PRO cameras when using the TrackIR software.

If you want to adjust the parameters you can use the unofficial/undocumented INPVector3 interface. You can use the Get methods to extract the defaults.

Pixel Width
Pixel Height
mm Width
mm Height
Focal Length

Birch,

Thanks so much for the reply. I'll look at that INPVector3 interface.

BTW... using Tsai's model, I got the following:

f = 2.849240 [mm]
kappa1 = 8.362079e-02 [1/mm^2]
Cx = 182.793037, Cy = 133.506785 [pixels]

Does that sound right?

Thanks,
Jeff

I am not familiar with the details of Tsai's calibration, but if you are using the stock C120 lens then the numbers aren't too far off (2.6 expected). The center X and Y coordinates seem reasonable as well.

hi everybody,
you can also use OpenCv Library for doing that simply.
http://opencvlibrary.sourceforge.net/

Hi,
Can someone from natural point just clarify the camera parameters for the TrackIR4 Pro camera?

Is it just the same as the FLEX:C120 camera in default configuration?

If I want to use INPVector3 to calculate the position of the vector clip, I'd need to update the focal point/height/width parameters right?

I do plan on getting one of the OptiTrack cameras just as soon as I've managed to get my software working with the trackir4 pro camera

Cheers,
Kyle

The pixel and width parameters for a C120 are the same as a TrackIR 4 Pro, but the focal length will not be the same since a different lens is used. The stock lens on a C120 has a focal length of 2.6mm, the stock lens on a TrackIR 4 Pro has a focal length of 2.45mm.