A couple months ago, my esteemed collegue Mary Gannon forwarded me an issue of a Machine Vision eNewsletter she gets from Pete Kepf, an AIA-Certified Vision Professional and consultant based in Louisville, Tennessee. I highly recommend his stuff: His advice to plant and design engineers on how to apply machine vision to get good results is very down to earth.
So, we invited him to visit the Design World offices in Cleveland to chat with us about top trends in Machine Vision … and here’s the conversation we had:
According to Kepf, two common causes of underperforming vision systems are:
Misapplication in the form of starting with the product instead of the problem. You know, if you have a hammer, everything looks like a nail.
• Unknown process inconsistencies.
Unknown process inconsistencies stem either from not knowing all the process variables or some unintended change to the process after system implementation.
When asked to compare major machine-vision brands — Cognex, Keyence, Sharp — Kepf points out that different platforms are suitable for different applications. So, he remains somewhat hardware agnostic.
Case in point: Here’s a part-sorting and product-monitoring application that Kepf setup using a Baumer VeriSens and a structured light source to collect height and feature data on a series of steel billets.