Beyond Manufacturing: Does machine vision’s future lie outside the factory floor?

Machine vision experts have made an industry by automating quality inspection and assembly operations in manufacturing plants around the world. The vast majority of traditional machine vision applications deal with the inspection of components for size, shape and other physical attributes that relate to the quality of the product. But there is a growing number of non-traditional, or non-manufacturing, applications insiders believe will lead to a fundamental change in the vision industry.

‘‘We define non-manufacturing vision applications as those that deal with more human-centric applications rather than industrial applications, such as medical life sciences and the automated reading of x-rays, for example; motion analysis, animation, transportation (such as adaptive cruise control or vehicle guidance systems), 3D modeling and more,’‘ explained Mike Melle sales manager for Basler Inc., the U.S. subsidiary of Basler AG.

‘‘We’re convinced that we need to spend more attention on the non-manufacturing areas of machine vision,’‘ added Basler AG’s CEO Dietmar Ley. ‘‘We’ve organized our business into two models: a systems business that produces turnkey systems for manufacturing, and a vision components business that makes cameras and other components that go into non-manufacturing applications.’‘

Industrial non-manufacturing: process control
As machine vision becomes more understood and the tools more robust, traditional industrial vision applications are solving new problems, such as process control. Printing applications and chemical plants are two examples of how one vision system can replace arrays of hardwired sensors to provide objective data for optimizing ongoing processes that do not generate products themselves.

The printing industry has had to depend on experienced press operators and/or expensive arrays of photoelectric sensors to make sure that the colors in the first newspaper to fly off the presses are the same as the colors on the last newspaper. To automate this process and reduce the complexity and system cost of the color control system, Applied Vision’s (Akron, Ohio) KromaKing™ color vision system offers the color sensitivity approaching that of a spectrophotometer while providing press operators with data they can understand. Color values are presented in a variety of common color spaces including Detal E, and Lab among others. ‘‘This is a paradigm shift in the industry using vision technology,’‘ said Applied Vision’s vice president of business development, Bud Patel. ‘‘[KromaKing] is able to get accurate, repeatable color measurements throughout the printing process.’‘

Similarly, Tri-Star Automation (Houston, Texas), an automation solutions provider (ASP) and distributor of DVT Corp. (Duluth, Georgia) cameras and vision components, was able to replace an expensive array of thermocouples and pyroelectric sensors monitoring open-flame burners in a chemical refinery with a single vision system. ‘‘Using multiple regions of interest (ROI), we were able to use one vision system to replace several sensors and combine it with a wireless transmitter to let a central control facility a mile away know when a pilot or burner had gone out by measuring the pixel intensities within the specific ROIs,’‘ explained Bill Athens, Tri-Star’s DVT application engineer.

Outside the factory: A whole new market
As the Applied Vision and Tri-Star applications show, when the cost of a vision system is equal to or lower than similarly discreet sensor solutions, then vision wins because of its flexibility and better analytical tools. But how will this success change the essence of the machine vision industry?

Convergence may be part of the answer. Convergence describes the combination of voice, video and data into a single digital environment. It is made possible by an ever-growing computing network that reaches deeper into every one of our lives each day – both at work and at home. Convergence has shown us that, if demand is strong enough, any technology can become a commodity – just look at the GPS and CDMA, both technologies developed by the military and now widely used by consumers.  If this paradigm also applies to imaging, then digital video and image processing could soon become commodities themselves.

To Basler’s Melle, digital sensors are already commodities. His company’s value add comes from optimizing digital cameras for high-volume OEM systems. Basler's VC division offers customers engineering expertise coupled with a selection of software and hardware based image algorithms that can be combined inside the camera to address the customer’s performance needs and cost requirements.

‘‘The customer already knows they need vision, and they want a camera that does just a few things and so they come to us,’‘ Melle said. ‘‘We’ve seen this in the medical, life sciences, animation, astronomy and many other non-industrial, non-traditional applications.’‘

According to Edmund Optics (Barrington, New Jersey) president and chief operating officer (COO), John Stack, machine vision libraries could soon find themselves on a ASIC chipset on every motherboard, not because people want to inspect their cookies as they come out of the oven or because industry wants to hook a camera up to a laptop and solve any inspection quandary, but because of a growing base of non-manufacturing imaging applications.

Automotive guidance. Self-serve check outs that work just as well with apples as they do with bar-coded diapers. Smart weapons. Package scanning and sorting for logistics. Biometrics. Service robotics and smart security.

‘‘When you look at these applications, these markets are much larger than factory automation. As it turns out, the big guys are all working on these ‘outside the box’ applications,’‘ Stack said.

Expertise versus R&D: Vision experts can benefit – big time!
And by ‘big guys,’ Edmund’s Stack doesn’t mean large vision companies such as Cognex, but Microsoft’s development of digital imagery sorting and recognition software, IBM’s ‘‘veggie vision’‘ used at grocery stories to identify apples from oranges; Intel’s ongoing development of open source computer vision libraries and Sarnoff Corp.’s vision for vehicles, also called intelligent transportation systems (ITS). Hewlett-Packard and NEC are other tech-giants entering the vision game.

‘‘With all these resources going into vision, the question becomes how long will it take for a technology to emerge that is truly disruptive to the machine vision industry,’‘ Stack asks. ‘‘What happens when Intel puts vision algorithms on a chip set, or CCD cameras dynamic range advances to the point where lighting and optics are not as important to an automated vision system as they are today,’‘

‘‘The first step is to go into their worlds – the HPs and IBMs,’‘ Stack asserts.

International conglomerates betting with massive R&D investments that far exceed what machine-vision-only companies can invest have the resources, but not the experience of AIA’s members when it comes to fielding commercial vision systems. Future success depends on marrying the two.

Experts suggest there are a few key ways that integrators and machine vision suppliers and supporting associations can prepare for the upcoming dawn of ubiquitous digital imaging and capitalize on the revolution when it happens. The first step, as Basler’s Ley suggests, is to embrace the world of the non-manufacturing or ‘‘outside the factory’‘ automated imaging. Next, build relationships with tech-giants by offering real-world expertise in building and field imaging systems.

By working with mass media, academics and associations and groups in related technology fields, machine vision companies and associations can better demonstrate their expertise to companies that traditionally may not be aware of the machine vision industry and what it takes to field commercial imaging systems.