Quantifying machine vision performance

It’s always seemed to me that if a vision system is performing a measurement function it should be treated the same as any other measurement device. In other words, it should be subject to periodic calibration and GR&R-type assessments. Yet, in my experience anyway, this is seldom the case.

One issue is that people who develop machine vision systems rarely come from the world of precision metrology. This means their understanding of concepts such as “precision” and “accuracy” is often somewhat confused, and that’s a pity because it’s important to know the difference.

I recently came across a great primer on this topic on the web site of Electronic Design magazine. Under “A Trip to The Gun Show; Or, An Analog Parable About Precision Versus Accuracy” (January 4th, 2011,) Dave Van Ess explains the difference between the two concepts, and perhaps more importantly, also addresses why you need to know.

If you are responsible for a vision system that performs any kind of measurement, you owe it to yourself to understand the difference between accuracy and precision.