Image and Measurement Drift

[gskt]

I have a metrology application where I am getting drifting measurements and the image appears to be changing position or getting larger and smaller. I am measuring the diameter of a ring. Both the camera and ring are in a fixed position and will see ambient temperatures (70-100 degF). I have been told that the change in temperature will affect the camera and will cause the measurement to change, but I don't understand why the whole image appears to move (very very slightly). I have tried different mountings, but still see the same effect.

[udka]

What camera you are using?
Generally industrial cameras should behave normal under camera specified temperature.
Possible to post images of ring? Normal and shifted?
How much variation you are getting in measurements?
How you are measuring ring Dia?

[BruceAmmons]

What lens are you using?

What is the scale of the image and the measurements?  Are you measuring on the micron level?

Bruce

[gskt]

I'm using an Allied Vision G-917. Attached are two immages that show a difference. It looks like the entire gasket and plate it is on has grown/shrunk.

For this example I am getting 0.005" variation. Overall I have seen up to 0.016" variation.

I am measuring the ring diameter by taking the perimeter and then converting it to the diameter (the ring isn't perfectly round so this gives a better reading)

[gskt]

I'm using a 25mm C-mount lense with manual iris and focus.

The measurements are taken on .001" level.

The resolution we are using is 2710 x 2710

[BruceAmmons]

The scale has definitely changed a slight amount.  Lining up the background in one corner, then comparing the background in the opposite corner shows several pixels change.

My guess is your focal distance has increased or decreased by a small amount.  Just enough to give you a few pixels of change in the scale.  Not much, but enough to give you a change in your measurements.

One possible solution would be investing in a telecentric lens.  These are designed to not change scale as the distance changes.

Another possible solution would be including calibration marks on the background.  You could always start your measurement by measuring the distance between two calibration points, then using that to scale your measurement results.

Bruce

[gskt]

Thanks Bruce. I'm thinking you are right about the focal distance.

For some of the images we do have calibration marks in the background. I went back through the images and reprocessed them and scaled the measurement results accordingly. The spread of measurements were smaller, but I still saw the same thing as before.

I would like to try a telecentric lense, but my issue here is cost. We are currently measuring objects up to 6" diameter and would like to go larger.

[BruceAmmons]

The way you measure your calibration marks can make a difference.  Can you measure them to sub-pixel accuracy?  Do they span a large distance (preferably most of the field of view)?  If you can only locate the calibration mark to the nearest pixel, you will still get variations in measurements.

I agree the telecentric lens becomes cost prohibitive when you get to 6" and larger.

Bruce

[gskt]

Thanks Bruce. Can you explain a little more about sup-pixel accuracy? I'm not sure if I can or not.

The calibration marks span almost the entire field of view (6").

[udka]

Subpixel accuracy calculation is described here: http://digital.ni.com/public.nsf/allkb/09B2BAFB6519A698862575B7007E3622
And here: http://zone.ni.com/reference/en-XX/help/372916T-01/nivisionconcepts/edge_detection_concepts/