Hello Guest it is July 10, 2020, 04:20:53 PM

Author Topic: Calibration  (Read 10025 times)

0 Members and 1 Guest are viewing this topic.

« on: June 19, 2008, 07:51:30 PM »

My probe is at the CNC Workshop in Illinois this week so I cannot do any serious testing until I get it back. However, I have given calibration some thought and I did do a few simple experiments during the brief time that I had it.

The present calibration scheme requires a square bar of known dimensions to be placed on the table parallel to the Y axis, and the camera and laser in the YZ plane and the target scan in the XY plane. No motion is used in the calibration so data is captured in a single position.

This static scheme does not allow for:
misorientation of the probe,
misorientation of the calibration bar,
inaccurate calibration bar dimensions,
radial distortion due to camera lens,
rectangular distortion due to non-square detector pixel layout,
keystone distortion due to laser and camera axes not perpendicular,
preferring a different arrangement of probe and target.

I would propose using a single point as a target. This could be the pointy end of a nail or pin or it could be the apex of a cone. If it is a cone then the apex angle would have to be sharp enough that the camera cannot see where the laser hits the cone except at the point. For example, in my setup at http://www.machsupport.com/forum/index.php/topic,5023.msg42913.html#msg42913, with the laser 30 degrees off the Z axis, the calibration target cone point would need to be less than 60 degrees.

I would jog until the point shows up near one corner of the screen then save the (X,Y,Z) coordinate from the machine axes and the (H,V) pixel horizontal and vertical coordinate from the CCD array where the center of the bright spot appears. I would do this for three corners to get three points as far apart as possible. This would create the data set:

(X1, Y1, Z1, H1, V1)
(X2, Y2, Z2, H2, V2)
(X3, Y3, Z3, H3, V3)

Three points are sufficient to define a plane so this is sufficient to define (H,V) points on the CCD plane in terms of the (X,Y,Z) points of the machine axes.

More than three points would permit a least squares fit which would be more accurate. Many more points, spread over the array, would allow for correction of distortion.

This scheme allows the probe to be installed in any orientation, even off axis orientations to get better data from odd shaped targets.

The probe could even be installed on the Z axis and pointed toward the front of the machine rather than down at the table. For this arrangement the point used for calibration would have to be setup so that it is pointed at the Z column. Then scanning the side of something would be possible.

Or side scan a target on a rotary table.

Or hang the probe on an articulating arm any way you wish.

Tom Hubin

Offline Chaoticone

  • *
  •  5,629 5,629
  • Precision Chaos
    • View Profile
Re: Calibration
« Reply #1 on: June 24, 2008, 11:12:57 PM »
Tom, get to play with it anymore since the workshop?

You really post some informative information.  :)

;D If you could see the things I have in my head, you would be laughing too. ;D

My guard dog is not what you need to worry about!
Re: Calibration
« Reply #2 on: June 25, 2008, 01:23:33 AM »
Hello Brett,

> Tom, get to play with it anymore since the workshop?

No. My video probe attended the CNC Workshop but I did not. With luck, I will get my probe back Thur night at our local amateur metalworking club (http://www.cams-club.org/) meeting.

> You really post some informative information.  :)

Thanx. In the 1980's I made my living designing non contact measurement equipment. Some used laser triangulation like these 3d video probes.

Others used trilateration by tracking a spherical glass cateye retroreflector about the size of a baseball. We could locate the cateye within 0.001 inch up to a range of 70 feet. We counted quarter wavelengths of red laser so we could have nailed the location within a wavelength of light if it were not for the mechanical tolerances of the gimbaled tracking equipment.

Tom Hubin
Re: Calibration
« Reply #3 on: September 02, 2008, 10:24:45 PM »
Ignore this post and attachments. I am just testing for a way to post MathCad equations.

Tom Hubin


Re: Calibration
« Reply #4 on: September 07, 2008, 11:17:41 PM »
I derived some equations to work with a calibration method using a single fixed point. This could be a common pin with the point up and clear of other surfaces.

I did the derivation using MathCad 2000. MathCad's native MCD files cannot be uploaded to the Mach Forum so I tried converting the file to HTM and to RTF. These also are not acceptable to Mach Forum but can be read into MS Word then saved as a DOC file. However, MS Word 97 and 2000 show the RTF and HTM files as garbled and saves the DOC file that way. So no luck there.

I tried copying to the clipbard, pasting into MS Paint, then saving as a GIF file. This seems to work ok. I am not sure how it will look on the Mach Forum or how it will print. Here goes.

Tom  Hubin
Re: Calibration
« Reply #5 on: May 31, 2009, 10:21:57 PM »
Attached here is the calibration math for Scan3d, the Open Source 3d video probe plugin posted at:   


The equations were derived using MathCad then copied and pasted to Microsoft Word for ease of sharing.

The method is a least squares fit using a single point target viewed in multiple positions. This method compensates for some lens distortion. Insufficient distortion compensation will be obvious if multipass scans do not align well. If multipass mismatching occurs then more terms will need to be added to the equations. 

Tom Hubin