Vision Validation: How Do You Know Your Vision System is Working Correctly?

Automated machine vision inspection is everywhere in today's world of manufacturing where human resources are at a premium and difficult inspections must be completed at a faster pace. With the integration of deep learning into these machine vision systems, how do you know that these systems are passing good parts and failing bad parts? Is using red rabbits or production managers making fake defects the way you test your systems?

Vision validation is a new concept for machine vision systems that you can use to test your inspection system to determine if its still operating within good parameters. This can be done by having a challenge set of good/bad images that can be run past the system that in theory should produce the correct results. It's not only a question of whether the vision system passes/fails correctly; you also want to know if someone has changed a parameter within your program that could fail a good product or pass a bad product. Vision validation is a new way to automate and routinely challenge/monitor your machine vision systems to verify that they are working correctly.

About the presenter
Eric Hershberger has over 24 years of machine vision integration and application experience. He can program robots, plc's and of course, cameras. Hershberger knows a few tips and tricks to make life easier for those difficult to solve vision projects that never have a good solution. He spends more time figuring out how to simplify the solution to make the support of the application easier. Hershberger loves the challenge of tough problems and the critical thinking required to solve them.