How to measure the thickness of quarter of a circle object in visual builder 2011(object like a curve , but the thikness of the curve in different edges point is not the same )

[MYounis]

Hi 

we have a smart camera NI 1742 and Vision Builder 2010 for evaluation.
We are building a system that can measure the thickness of the lower welded double
seam of an aluminium food box . by lighting system the welded double seam shown in
dark colour .

A defect box have a bigger thickness than the max thickness that suppose to be.
The box have a 4 corners , each corner is like a curve (quarter of a circle), How to measure the thickness of quarter of a circle object in visual builder 2011(object like a curve , but the thikness of the curve in different edges point is not the same )
we could not find a good and a fast way to do it by using vision builder 2010.

I attached 3 pics . one with a description and one Pass and one Fail .

Thank you,

[MYounis]

Please help me

[BruceAmmons]

If you can convert your black shadow to a binary object, you could use the Distance or Danielsson functions, which tell you how far from the nearest edge a pixel is.  The max value will tell you the thickest section.

You might want to work on your lighting.  The glare in the bottom of the can might make it difficult to do the analysis.  Ideally, light coming in at a 45 degree angle on all four sides would work well.  It would reflect a few times and exit at about 45 degrees.  The only light that would reach the camera would be the diffuse reflections.  Not sure what the rounded corners would do, though.  Another possibility would be a very large diffuse light.

Bruce

[Matt_McLaughlin]

Hi MYounis,

If the position of the images is going to be relatively constant, a simple, but not highly accurate option would be to set up a series of find edges and geometry (distance) steps. You could find edges along lines running perpendicular to the curve of the object and then use  a geometry step to measure the distance between the inner and outer edges that will be detected. It would be important to add a relatively large number of these edge detections especially in areas that you would expect to find imperfections.

Again, this method will not be ideally accurate, but if the failures are expected to be on the level of the image you included in you previous post, it would likely provide a simple solution.

I hope that's helpful!