Hi guys,

I'm using the 3 kinect setup which makes the groundplane auto-detected based on the depth stream.

Well, is there a way to adjust the groundplane (afterwards)? The problem I'm struggling with is that my real character stands on a plate (dimensions 150cm x 60cm, height above real ground ~15cm). That issue makes my virtual character always stay above the detected groundplane.

It seems that "Recenter Coordinate System On Character" only makes changes to the orientation and position (x-z-plane, if y is up-axis) in respect to the global system. If I change the virtual character's position along y-axis (Up-Vector) the groundplane still sits at its old location.

Any ideas how to fix it?

Regards
Andy

Hi
There is no adjustments for groundplane for depth sensors calibration right in Mocap Studio.
But you can manually adjust the height of the cameras in a scene file obtained as a result of calibration. It is a human-readable (and editable) XML file. Decrease the Y (second) value of Translation vector for each of the cameras by 0.15 (15 cm) - that should do the trick.

Awesome. Thanks a lot for that hint. But don't I add any errors (position, angular) if I do that adjustment?

Also, two of the kinects - the outer ones (i.e. Kinect 1 & Kinect 3) already show me the correct height values (which are 155cm in reality). It's only Kinect 2 thats height value isn't calculated correctly.

My setup is as following:


_______________Kinect 2______________

......................Plate.........................
......................Plate.........................
......................Plate.........................
Kinect 3______________________Kinect 1

The relative positions of cameras do not change when you "move" them all synchronously. It is simply the shift of coordinates origin.
Thus, the accuracy of tracking will not suffer.

What do you mean by correct height - over the floor or over the plate?

Of course, it is better to have a flat ground surface, at least for calibration. Either by (temporarily) dismounting the plate or by raising the surroundings to the plate's height with some mats.

You may also provide me one of your calibration videos to get better understanding of what's happening and test the calibration. You can use some file-sharing service like Google Drive or Dropbox for that.

Well, every of the three Kinects in the room is mounted at a height of 155cm.
After calibration IPI MocapStudio tells me that Kinect 1 and Kinect 3 are at that height. Only Kinect 2 is calculated to be at a higher location (i.e. 172cm).

I guess the calibration takes the depth information separately for each kinect. Kinect 1 and Kinect 3 have a good overview of the room and so the plate is only a small part within the captured area. -> Height of Kinect 1 & 3 is calculated based on the ground of the room.

The plate is taking up far more space in Kinect 2. I guess when calculating the height location of Kinect 2 the calibration "thinks" that the plate corresponds to the real ground and takes that into account for calculating the Kinect position.

Temporarily removing the plate? Well, that's not an option, it weighs roughly 200 lbs. ;-)


The calibration error was 0.056 (position), but while running the calibration there were a lot of tracked points marked yellow and grey. Is there kind of a guide how to avoid those? The maglite seems to be visible in every frame - but is not detected like it should be. The room should be the perfect place for calibration, it's used for a Virtual Reality System (270cm x 200cm).

Also, Kinect 1 and Kinect 3 are Kinect for Windows, Kinect 2 is Kinect for XBox. Though I play around with the kinect settings (gain, exposure) of the Kinect for Windows the maglite always has a much better visibility in the Kinect for XBox.

I'll do another calibration and upload it to Dropbox cause the current one is about 3,2gb.


Thanks a lot for your time in helping me.

Regards
Andy