Why machine vision is difficult

The basic problem is that computers are dumb: they’re machines that just execute lines of code. This means that when you develop a vision/inspection algorithm you need to anticipate every possible condition the system might encounter. To put it another way, the challenge is to minimize the number of variables in the image. The more you can fix, the less complex, and hence more robust, your application will be.

Ned Lecky of Lecky Integration discusses this subject in “See the Nail. Hit the Nail,” on the Control Design web site. I recommend you find out what Ned has to say, for two reasons:

First, as the father of the Sherlock package, he’s one of the “Grand Old Men” of machine vision, and he knows what he’s talking about.

And second, an understanding of what makes machine vision so difficult will help you to be more successful. (Well, more successful with machine vision. I doubt it will make a difference in other areas of your life, but if it does, please let me know.)