Calibrating a HD camera to a Kinect Depth

edit: arrrrgh, should’ve read the thread completely before posting mindless babblings… :/

Hi Gundorf / Elliot.

Did you manage to get this working?

Just having a play with rgbd toolkit now

edit - anyone know if it’s possible to switch off osd and view hdmi video without recording using a nikon slr (d3100)?

Shoulda got a canon :)

I’m pretty in rush right now but I will post some patches soon including great elliot woods advices and techniques !


Gotta love Elliot’s advices and techniques :)

re: HD camera - just been testing the Logitech c920 HD Pro Webcam and really impress with the results. Getting 720p@30fps with hardly any latency. Just have to put a single frame texture delay on the kinect textures and they are both synced. Also 1080p@30fps working requiring only 2 frame texture delay.

Not solved the calibration process but getting on it slowly. Done a cheap hack version that gives OK results.

but yeah good camera.

@xd_nitro: that’s my review of its older sibling (C910)

developments with this problem:

the package in the video can’t be distributed as it contains VVVV (which afaik, i’m not allowed to distribute)

expect it in next OpenCV release or contact if you want the package.

hi gundorf, i’m trying to get a rgbd+projectorCalibration to work. could you give me some advice?

got elliots rgbd-patch (thank you!) to work and even managed to run that at the same time with calibratieProjector.

now i think i have to combine the transformation from rgbd with the transforms of the calibrateProjector-Patch, but, so far, i have no idea how :)

Hello guys,

i have a few questions about the RGB-D Toolkit and your solutions for vvvv. Most of my question may seem a bit dumb :)

RGB-D Toolkit:

  1. to get the RGB-D Toolkit to run on windows, do i need to set up an openframeworks installation? i cant find any windows download, only the project files on github.

elliotwoods Solution in vvvv:

  1. Are the example patches from your youtube video available somewhere, or am i able to rebuild this with the standart openCV modules? Any help on this? :)

  2. Would it be possible to adjust and calibrate the camera+kinect both once on a fixed mount and apply the transformation on any footage filmed with this setup? In other words: Would i need to calibrate the cameras each time the scene changes, or does one calibration work in every scene?

Help and more infos on this would be great :) Im currently using a svideo input for the calibration since the blackmagic shuttle is not working on my laptop, but i assume applying the same transformation on the same aspectRatio footage videofiles would give the same results? :)

I’m using Elliot’s toolkit. And I’ve done everything he says in the video.
I can’t understand why when I it “do”, I see all black.
Any help? Pleeeease :)

since you’re not seeing a camera frustum at all (the preview of the rgb camera’s perspective in the world scene), then there’s something seriously wrong with the calibration.
did you try the files from ?

Thanks for the link Eliot, i have been looking for this package for a while too.
Only problem though is that the moment it activates the Kinect, vvvv freezes and i have to kill it with the task manager. When i kill it however it fails to unload/release/disengage the Kinect. After this if i unplug the Kinect from the USB cable in order to reset it the computer fails o re detect it until i restart the system.
The same result is on 2 different computers and both OpenCV pack and the newer Imagepack.

Anyone having any idea about this ?

@elliot I set the drop value of KinectWorldProjectedTexture below 1.00 and it worked. (I don’t know why :) )

@synth do you start the kinect with a bang or a toggle? I had the same problem when I used a toggle. You should start it with a bang

I decided to do a full reinstall of both computers and now it works.
Didn’t have the time to test caribration yet but at least it is not freezing vvvv now.
As you said the default activation was a toggle.

Only 1 hing though when using the new imagepack the kinect nodes are red but if i add the old OpenCV contribution everything works fine.

Getting confused with all the contribution libraries … some nodes are in the older pack and some in the newer …
time will teach me i hope :)

Anybody knows if it is possible to use this calibration not onto the mesh from kinect, but on the plain depth view?

hello all,

am trying to achieve the same thing but using opencv and wind sdk. can some one guide me through mapping the external RGB to depth image?

i have the intrinsic and extrinsic parameters from RGBD toolkit. just need to know how to apply it.


Any news on that project elliot ?