Dawson notes that fundamentally, machine vision tasks are difficult. “You’re attempting to replace human vision. So first, we have to ‘cheat’—simplifying problems, positioning parts and limiting the number of parts.” What’s created is primarily an inspection tool, “looking for defects, things that don’t match some template,” Dawson explains.
Still, why is machine vision installation difficult? “You’ve got other issues about lighting, cameras and the like,” Dawson says. “Also, you’ve either got to have expert knowledge or a vendor’s products with built-in intelligence.”
ipd—which stands for intelligent products—addresses the latter through a “vision-engineer-in-a-box.” As Dawson notes, “We spend a lot of time and effort on design of the human interface.” Typically, there’s a design-time interface and also a run-time interface, which is exposed to an operator. He notes that in ipd’s products, design and operator interfaces are the same. “The end-user [designer] simply specifies what should be exposed to the operator.”
Driving ipd’s interface focus, Dawson says, is this principle: “If you can’t access things in terms with which you’re familiar, then the technology is useless.” His “golden rules” for making interface design user-friendly? “Don’t have a lot of learning curve. Don’t make me think. Don’t leave the user wondering what to do next. Tell it to me in terms I’m familiar with,” he says. “Finally, make the interface in terms with which end-users are familiar.”
Even so, John Keating thinks it’s “extraordinarily difficult” to make an interface simple. The goal is making things obvious, states Keating, who is the Checker vision system product manager for Cognex Inc. (www.cognex.com), a Natick, Mass., machine vision vendor. One way Cognex does that is by avoiding
wizard-based installation and configuration systems. But why no wizard? “You don’t know where you are in the process, if you’ve got to go back to the wizard. That’s frustrating—and the user doesn’t learn anything by just clicking ‘Next,’ ” Keating believes.
And just as Dawson points out the importance of simple language and directions, Keating notes that his objective is to convey complex concepts with clear, concise explanations. To achieve that, Cognex simplified Checker’s installation-and-configuration schemes. “The steps are named very obviously. If I want to change where I find the part, I go the place where it says ‘Find my part’ and make changes.”
For physical installation, Keating notes that the Checker system, which he calls his company’s “easiest product to use,” comes with a bag of camera lenses. That helps eliminate the trial-and-error that end-users typically face when trying to find the correct distance from sensor to object. Also eliminated is the need for an external photoelectric sensor, which means that once the system is powered up, it immediately starts taking pictures. Set-up is also non-intrusive, Keating says, meaning no stopped production.
Keating mentions yet another installation and configuration simplifier: “not having a boatload of controls, options and switches to provide flexibility.” While that may sound counterintuitive in today’s bells-and-whistles, ultra-flexible world, the two main things that end-users want from a vision system are finding, and then inspecting the part, says Keating. “And in a nutshell, that means making things in the [vision] product familiar to the users. It gives them the feel of something that they already understand how to use.” That’s smart.
C. Kenna Amos, firstname.lastname@example.org, is an Automation World Contributing Editor.