Tuesday, November 26, 2013

Finding cracks


Defect detection is perhaps the toughest challenge in machine vision, and cracks are the defect I find hardest of all. Dalsa’s Ben Dawson offers some excellent advice on image processing s to highlight cracks in “Detecting of Cracks and Scratches” [sic] published in Quality Magazine, November 14th, 2013, but that supposes the crack is visible in your image.

The problem is, cracks are often extremely thin, even though they might be many millimeters long. If the crack has to be three pixels wide to be detectable, that means you need a very high resolution. The only alternative I know of is to use some kind of fluorescent penetrant to highlight the crack. Failing that, you might be looking at ultrasound or eddy current.

So, good article Ben, but you overlooked how to make that crack appear in the image in the first place.

No comments: