Sunday, January 5, 2014

Study this chart

I dislike lifting images from the internet: it seems like a breach of copyright. But just occasionally one sees something so important it has to be copied.

So it goes with the chart below. This comes from “Turnkey Machine Vision Systems”, written by Ross Rawlings of machine vision integrator Radix Controls and published in Quality Magazine December 10th, 2013.

The sharp-eyed will see it explains the reasons for false rejects. And more importantly, it makes the point that if a system isn’t rejecting good parts it may well be passing bad ones.

Now of course, the really clever integrator/programmer ensures there’s sufficient difference between good and bad that the distributions are separate. In presence/absence applications that’s not too hard to achieve. It’s harder though in gauging, where you have some measurement uncertainty, and well-neigh impossible in surface flaw detection. So study the chart well and be prepared to discuss it with your customers and managers.

Two side notes before I take my leave:
  1. I would love to see some serious research done into the mathematics of false rejects. If anyone knows of any, please let me know.
  2. A personal note to author Ross: I try to incorporate a little copyright notice into any graphics I produce. You might like to do the same.

1 comment:

Anonymous said...

It does not only seem to be a breach of copyright, it *is* a breach of copyright.

Any graphic on the internet is copyrighted regardless of whether there is a copyright notice (unless copyright is relinquished explicitely to put it in public domain). More commonly the author might simply grant everyone permission to use it freely e.g. via a CC license, in which case he still retains copyright / ownership.

If neither is the case author needs to be before using image.contacted