Re: [CAD_CAM_EDM_DRO] Camera-laser scanning

ibewgypsie wrote:

>
> http://img.photobucket.com/albums/v85/ibewgypsie/Sscn0998.jpg
>
> This is how it works.. fuzzy picture, sorry.. You can get a general
> ideal thou. Looks at that height and angle the resolution is approx
> 640 pixels for 3" As good as some probes but not as good as I'd like..
> I can of course raise the table untill I got 1" FOV then reangle the
> camera.
>
> Like others said, this is not as precise as a probe or distance scanner.

*****************

Hello ibewgypsie,

Beware the Scheimpflug Condition! Details patented about 1904.

I'll give you the short version then you can do a google search on
"Scheimpflug: for technical details.

The object for your camera is the plane containing the laser. That plane
is to be imaged onto the camera detector or CCD plane. The problem is
that the laser plane is not parallel to the lens plane. This results in
a tilted image plane. The CCD array should be tilted for sharpest focus
over the useful laser range.

The math is:

tan(beta) = m * tan(alpha)

where:
m is the magnification along the optical axis.
alpha is the tilt of the object plane
beta is the tilt of the image plan.

If the camera has a very small aperture then the depth of field is large
and the object tilt may not matter much. But if a large aperture lens
(low fNumber) is used then a great deal of blurring will occur for
points away from the center of the useful laser range.

Another saving point is that the magnification is usually a fraction for
laser triangulation. This means that the image plane tilt is much less
than the object plane tilt.

It is difficult to tilt the CCD array in a typical camera. In the late
1980's I designed laser triangulation gages for a living. We designed
and built the camera from scratch so I could specify and build the CCD
tilt according to the math and verify with ray tracing software.

Also, tilting the image this way results in a nonlinear mapping from
object to image. Like the human eye, resolution will be best for points
closest to the camera. Resolution will be worse for poitns far from the
camera.

There are optical methods that will allow a linear mapping (constant
magnification) from object plane to image plane but you won't find that
in a common camera. Ask if you want the details.

Tom Hubin
thubin@...