Surveyor stereo vision calibration

[Lucas]

(already posted most of this as a googlecode comment and accidently sent it just to one individual rather than the list, sorry for all the duplication)

For calibration, does every dot in the pattern need to be visible, or should the pattern fill the screen with dots going off the edges?

Should the red dot be in the center?

Does the pattern have to be straight on, so that the calibration pattern surface is perpendicular to a line going from the camera to the surface?

I'm trying to calibrate and so far the only time anything has happened was when I held the camera very close, and it then it was 'corrected' to an extremely distorted view- this is using the svn version from yesterday.

Also I was able to build and run this with Visual C# 2008, though after a few minutes of connection something goes wrong and the orange led toggles off and an adjacent orange led turns on on the camera boards. 


Thanks,

Lucas

--
http://binarymillenium.com

[motters]

> For calibration, does every dot in the pattern need to be visible, or
> should the pattern fill the screen with dots going off the edges?

The pattern should fill the screen so that no other background objects
are visible. If some dots are off the screen because the pattern is
close to the camera that doesn't matter too much. You can experiment
to find the ideal distance by moving the pattern towards or away from
the camera.

> Should the red dot be in the center?

Yes, somewhere near the centre of the field of view.

> Does the pattern have to be straight on, so that the calibration
> pattern surface is perpendicular to a line going from the camera to
> the surface?

Yes.

> I'm trying to calibrate and so far the only time anything has happened

> was when I held the pattern very close to the camera, and it then it

> was 'corrected' to an extremely distorted view- this is using the svn
> version from yesterday.

The calibration can fail sometimes, but this only happens rarely in my
experience. If this happens it's quite easy to try again.


> Also, the stereosurveyor project successfully builds and runs from
> Visual C# 2008 Express Edition with some minor warnings about unused
> variables, but sometimes after a minute or two of operation
> (attempting to do the calibration) something goes wrong, a different
> orange led on the camera boards lights up (and the orange led closest
> to the red led goes off), and I have to power cycle the svs in order
> to reconnect.

This sounds like it might be a firmware bug. I only have an old
pre-production version of the SVS running an old version of the
firmware. I did have the SVS communications lock up on one occasion,
but it turned out to be due to a faulty matchport.

I havn't used VS2008 at all, only VS2005 and MonoDevelop, so it's good
to hear that it still works on the newer IDE.