[sldev] Anyone here with OpenCV experience?

[Philip Rosedale]

Have you seen this video...  http://www.youtube.com/watch?v=r7Gn2TyEyHw

It is a cheap logitech camera tracking a head and head rotation really 
well (which you can see by the registration quality of the video overlay).

This is what we want - simple head position and movement tracking from a 
camera.  I suppose we could just use the logitech API but that would 
restrict us to their cameras.  I believe that the code they ship with 
their cameras is based on OpenCV?  Hence the question. 

Philip

Jan Ciger wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Philip Rosedale wrote:
>   
>> Has anyone here worked with camera-based gesture recognition before?  
>> How about OpenCV?  Is OpenCV the best package for extracting basic head 
>> position/gesture information from a camera image stream?  Merov and I 
>> are pondering a Snowglobe project to detect head motion from simple 
>> cameras and connect it to the SL Viewer. 
>>     
>
> I have done something like that. First of all, OpenCV is does not do any
> gesture recognition. It is an image processing library and you need to
> know quite a bit about computer vision to be able to use it
> meaningfully. I have found this book indispensable for that:
> http://oreilly.com/catalog/9780596516130/ , plus a good basic computer
> vision textbook.
>
> If you want to track the orientation of the head with a webcam, that is
> doable - NaturalPoint is selling infrared webcams that track either a
> single point or a simple rig on your head and translate that to mouse
> motion. That is easy enough to do. If that is all you want to do
> (translate head motion to 2DOF mouse-like signal), OpenCV is even an
> overkill - capture image, threshold it to extract the marker and measure
> where it is compared to the initial position. Then use that difference
> to steer camera.
>
> The main issue you will have will be a cross-platform video capture
> library. I have yet to find one that is not too buggy and actually works
> without doing crazy stuff. OpenCV has some code for capturing video, but
> that is mainly for debugging and shouldn't be used for production (it is
> *really* buggy and slow).
>
> On the other hand, if you want to do something more complex - e.g. full
> 6DOF tracking, OpenCV is a good starting point. For example, I have the
> Minoru 3D webcam (http://www.minoru3d.com/) connected to a 3D tracker on
> Linux that allows me to track both the position and orientation in space
> - - like a 3D mouse. The tracker is made using OpenCV. However,
> calibrating that is not something for an average Joe to do ...
>
> Regards,
>
> Jan
>
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.9 (GNU/Linux)
> Comment: Using GnuPG with Mandriva - http://enigmail.mozdev.org
>
> iD8DBQFKFbPZn11XseNj94gRAmdkAJ9fbD8bpCVqK3KtJ1FTunFVSmVMTgCfRuvB
> VABbYQzKrgfLkcCI8i8iclk=
> =kMun
> -----END PGP SIGNATURE-----
>   
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.secondlife.com/pipermail/sldev/attachments/20090521/c7cc42c5/attachment.htm