The limits to growth

Old codgers like me might remember that as the title of a doomsday scenario book about how things can’t go on as they are. Looking back, it would appear the authors forgot that everything changes, all the time. And how is this relevant to my theme of “machine vision”?

Well I was just reading about some pontificating put forth by Paul Kellett of the AIA. He’s projecting a steady expansion of the MV industry, on the basis that “The number of applications to which machine vision is being applied is growing all the time,”. (Imaging and Machine Vision Europe, Dec 2007/Jan2008.) I’m not sure I agree.

I believe that machine vision is being held back by the need to develop a solution to every problem. This means that a lot of smart people are needed to figure out how to put vision to work, (if you’re reading this you probably know that vision ain’t easy,) and that makes it an expensive technology to deploy. Sure, hardware costs are coming down and capabilities (such a resolution) are going up, so we can do more with less every year, but the limiting factor is in figuring out how to do more with less.

Dalsa now have a 12k linescan camera. That’s great, and it may even help me solve some imaging problems, but I’ll still need a lot of time and effort to develop and implement a new web inspection solution. As anyone in the integration field knows, you only make money on the repeats. One-off systems are often money losers. But in the real world that you and I inhabit there aren’t that many replication opportunities, so we’re forced into doing lots of time-consuming development work. And that’s what will hold back the growth of this industry.

What’s the solution? I believe that vision software has to get smarter, much smarter, ideally to the point where we train it the same way we train MaryJo in the plant; by showing it good and bad and relying on it to learn. I’ll return to this theme on future occasions.