Hi,

I've post my problem in here long time ago. But my problem haven't 
solved yet. Here is the revision of my problem.

I'm doing position tracking with some marker base system (using computer 
vision technique). This system is finished and I'm going to use it as a 
base of an application which I'm going to develop. (It is about 
augmented reality)

What I've done so far is the simulation tracking and the real world 
tracking. The result of real world tracking is very accurate, with about 
+/-3cm deviation. But for the simulation, though the planar position (x 
and y) is very accurate, but the distance (z) is very inaccurate, with 
the average deviation of about 40cm. Which is very unacceptable.

For now, I'm going to vary the focal length of the camera to see if it 
is related, but seems it is not the source of error. I've read from this 
newsgroup that the pov-ray unit is not related to the real world unit 
(centimeter or meter). Would it be the source of the error?

And also, reply from the previous post suggest that I've distorded the 
output image. I've double check it and I'm not distorting it.

So anyone has any clue on it?

Here is my camera definition, which is simulating the real camera in the 
real world tracking:

camera
{
    location <61.5, 74.55, -200>
    direction <0, 0, 1.7526>
    look_at <61.5, 74.55, 0>
}

As this is a square pixel camera, and with normal 4/3 aspect ratio, so 
the image is rendered with the dimension 384x288 (both resolution are 
the same).

P.S.: markers are texture mapped into the virtual scene.


cheers
Colin

Post a reply to this message

> For now, I'm going to vary the focal length of the camera to see if it
> is related, but seems it is not the source of error. I've read from this
> newsgroup that the pov-ray unit is not related to the real world unit
> (centimeter or meter). Would it be the source of the error?

Though a POV-Unit is not "standardized" in any way to be either cm, or
meter, or mile, or whatever, this doesn't mean that the usual physics don't
apply (as in parallax effects, angles, distances, etc). It's just that you
can't tag unit = cm. Just think about it, even cm is just an arbitrary
dimension someone has thought up someday.

So it just depends on your scene-file if unit=cm or unit=meter or unit=?.
That just means you've got to multiply the distances with some factors.

As for the 40cm deviation: is that only happening on the POV-Ray images?
Could you post a comparison image of photo vs Pov-Ray image with markers to
p.b.i?

-- 
"Tim Nikias v2.0"
Homepage: <http://www.nolights.de>

Post a reply to this message

Tim Nikias wrote:
>>For now, I'm going to vary the focal length of the camera to see if it
>>is related, but seems it is not the source of error. I've read from this
>>newsgroup that the pov-ray unit is not related to the real world unit
>>(centimeter or meter). Would it be the source of the error?
> 
> 
> Though a POV-Unit is not "standardized" in any way to be either cm, or
> meter, or mile, or whatever, this doesn't mean that the usual physics don't
> apply (as in parallax effects, angles, distances, etc). It's just that you
> can't tag unit = cm. Just think about it, even cm is just an arbitrary
> dimension someone has thought up someday.
> 
> So it just depends on your scene-file if unit=cm or unit=meter or unit=?.
> That just means you've got to multiply the distances with some factors.
> 
> As for the 40cm deviation: is that only happening on the POV-Ray images?
> Could you post a comparison image of photo vs Pov-Ray image with markers to
> p.b.i?
> 

Thanks for the comment on my problem. But just a silly question...
what's p.b.i???

Post a reply to this message

news:4209fb22$1@news.povray.org...
>
> Thanks for the comment on my problem. But just a silly question...
> what's p.b.i???
>

the newgroup povray.binaries.images on this server

Marc

Post a reply to this message