Hi. David Wennström <[EMAIL PROTECTED]>:
> > Yes, it is a long time. BTW: If you do head tracking you can often get > > away with 5DOF tracking, since you are probably only interested in the > > two eyes. > > Which DOF can I skip? The "roll" angle? I am not sure which is the "roll" angle, but if we assume that your eyes lie on the X axis (head coordinate system), then you can just forget rotation about the X axis. Now your markers might be slightly off the X-axis and theoretically ignoring that rotation causes an error unless also the markers are on the X-axis. How-ever, most tracking systems are so inaccurate anyhow that nobody really cares about this ;-) We used to just assume that eyes are a bit beneath the markers (which were attached above the stereo glasses) and things went fine. Not hifi, but good enough for the purpose. > > > I am also very interested in links and projects that I can read about 3D > > > tracking system like this in general and particularly head tracking. > > > > > Here is link to one (that I wrote). I think most would be hidden in all > > kinds of SW systems. > > > > > http://csdl2.computer.org/persagen/DLAbsToc.jsp?resourcePath=/dl/proceedings/&toc=comp/proceedings/vr/2006/0224/00/0224toc.xml&DOI=10.1109/VR.2006.139 > > > > Thank you, I'll look into it. I can dig up the pdf if you cannot get from that site. > > We used Apple iSight cameras which were certainly cheap enough (but you > > cannot lock the white balance...) > > > > The unibrain board cameras would be very cheap, but then you are down to > > 30 fps again... > > > > It's not so much the fps but the latency from picture taken to getting it to > the PC. Currently I will be stuck with the SPC900NC camera - trying to get it > a little bit faster... :) Yes. There is data transmission latency and then many (most?) consumer cameras do frame integration to reduce noise (and increase effective latency), which of course is the last thing you want... -- Tommi _______________________________________________ pwc mailing list [email protected] http://lists.saillard.org/mailman/listinfo/pwc
