Vision System Calibrations or lack of???

Hi guys,

I have another question for you...not scantimes this time but vision systems.....Arrrggghhh!

Let me explain...

I have an Epson robot that places an rectangular pattern on a plastic body prior to silicon base components being placed upon it. If one was to draw a box around the pattern it would measure approximately 25mm wide and 40mm long. The adhesive is a continuous bead whose thickness on the black plastic body should be 0.4mm all around the pattern.

A vision system measures the bead pattern 10 seconds after it has been dispensed onto the plastic body. The vision system consists of a light ring approximately 20mm above the body with an inner and outer LED ring so the light can be varied across the plastic body. Attached to the bottom of the ring is a polariser. Approximately 100mm above this is a camera with a polariser as well(Melles Griot camera).

There is no problem calibrating the vision system so that the vision system can match images to real world co-ordinates...simple and extremely robust. However, when it comes to measuring the bead thickness, the whole system becomes very flakey. By varying the power of the two LED rings one can vary the thickness of the bead (as determined by the vision system). Therefore, a bead that might actually be 0.5mm wide can be made to read as 0.3mm or 0.6mm on the vision system depending on how you vary the light.

The adhesive is white, the pen body is black...how do I set my system up so that I know that when I put a pen under the camera I know that the results that come back are true....I can't measure the bead and then place it under the camera since it settles over time and a few seconds can make a huge difference!

Any further details just let me know....I don't know have I explained myself well enough!

I am not to sure if this idea will work exactly with your application but it is one method that i have used in measuring app's.

May need to adjust your threshold

The fact that the width reported by the vision system varies with light intensity might indicate that the threshold adjustments need to be tweaked.

On the cameras I use (Cognex), the edge tools have a number of parameters to establish the switching points (dark->light, light->dark, threshold, etc). Ideally, your transitions from dark to light should be very large and abrupt, providing a very repeatable reading. A modest change of light intensity would still produce essentially the same threshold crossing and thus the same measurement.

Suppose, however, that your transitions were NOT so large or abrupt. Maybe the peak light intensity is close to the threshhold setpoint. Then a change in light intensity would cross the threshold at a different location, giving a different measurement. If you have the programming software, you should be able to view a display of the threshold point of each tool.

SOLUTION : You must increase the contrast between the dark and light portions of your image. 1)Open the lens 2)increase the camera exposure time 3)increase the camera gain. If you are able to display what the camera is seeing, you should be able to tell when you get it "right".

Once one or more of these are done, then you should have more "excess gain" (to borrow from the photoeye world). This should greatly reduce the effect of light variation.

Adhesive bead measurment

The laser light method should work well.
Another method that will give you consistant results is to use infrared lighting and filter.
With any lighting a regulated power supply is a must. IR lighting is a bit more expensive but very stable over time and almost immune to ambient lighting.

ControlLogix,

You may test several different light sources until you get the most accurate and stable edge measurement. Especially with white adhesives beads an ultraviolet light system should provide perfectly contrasted results.

The laser solution can give good results by counting pixels or subpixels in a fixed measurement window in which the complete laser ray provide some predetermined values.

Laurent