I’ll mostly blog about the computational machinery and algorithms of natural and machine vision. I can’t go into as much detail on algorithms as Steve Eddins does in his blog on image processing but perhaps this is a relief. I’ll discuss our products when appropriate, but no snarky comments about our competition that could bring on a plague of lawyers. For witty and insightful comments on the business and personalities of the machine vision industry, I recommend B Grey’s excellent blog.
Are you confused about the different names used to describe machine vision technology, as examples “image processing” or “image analysis”? Or perhaps you never thought about it. OK, it’s not so important, but knowing what these different names represent can help you specify what technology you need for a vision task. And you might impress your colleagues.
Names summarize a set of properties. In our case, a convenient way to reference different schools of thought on how to turn patterns of light or other radiation into information. Brand names also make it easy to remember a product, be it commercial or academic. But enough Whorfian ruminations – bring on the names!
Machine vision is the automation of vision tasks that are too repetitive, fast, precise or boring for humans to do. Its essence is converting patterns of light into measurements and decisions:
- Is the fuel injector the right diameter?
- Is the biscuit burned or not?
In this Venn diagram, machine vision contains names of some technologies it uses. “Computer vision” mostly surrounds machine vision to suggest that machine vision “inherits” many techniques from it. I don’t know who came up with the name “machine vision” – perhaps the Brits did.
Computer vision is the academic relative of machine vision.
Computer vision is concerned with discovering general vision principles, such as how to determine an object’s structure from its movement, and with difficult and under-constrained vision problems such as automating breast cancer detection. Machine vision benefits from this research and computer vision benefits from improvements in machine vision technology, such as better cameras from Teledyne DALSA (of course). Some academics look down on machine vision as “just engineering” but I think the two fields are complementary. You might hire a computer vision “scientist” for a challenging vision problem, but most industrial vision tasks are best tackled by a machine vision “engineer”.
Well we are out of time, so let’s continue this discussion next blog.