Will machines one day be able to see just as humans do, as popular science is wont to desire? If you're a manufacturer, it probably doesn't matter.
"Machines don't really "~see' anything in the human, anthropomorphic sense," said John Parkinson, affiliate partner with Waterstone Management Group, a Chicago-based advisory firm focused on serving the technology sector. "What we have today," he told ThomasNet News in an interview, "is a variety of image-based analytical and sense-and-response systems."
Machine vision (MV) is very different from human vision and has its limitations. But the fact is that MV systems are constantly improving through incremental innovation and now able to perform tasks that are impossible for the human brain-eye complex, however marvelous that natural capability might be.
Microscan Systems, a supplier of MV technologies based in Renton, Wash., defines machine vision as "the automatic extraction of information from digital images." MV systems typically incorporate a camera and lighting to illuminate the part being inspected, along with computing systems for image processing and communications systems to integrate the vision function into the larger shop-floor production environment. MV suppliers are increasingly able to incorporate many of these functions into single-unit "smart" cameras.
In a basic sort of MV configuration, a steady stream of manufactured objects is exposed to image-capture as they pass through the field of a digital camera mounted at one point along the line. The image of each object is processed and compared against specifications to make sure it meets criteria and is free of defects.
Microscan Systems classifies the four most common functions of MV as:
1. Measurement of parts
2. Counting the number or features of parts
3. Location, or reporting the position and orientation of a part or identifying patterns on its surface
4. Decoding of symbols such as barcodes and data-matrix codes or of text, employing OCR.
Manufacturers employing machine vision are benefiting from improvements in cameras, optics, lighting, scanning technology, and vision software. One important area of innovation is 3D machine vision, which is seeing adoption in such areas as robotic control and bin picking.
"Where it's important to know the color, the size, the positioning of a part, where there's a label that has to be read, those kinds of activities, that's where machine vision offers value," said Parkinson. "A product line goes fast, and a vision system can see things at a speed that a human just couldn't. So in circumstances where you need pattern-recognition, shape recognition, confirming that "~this is the right thing,' you can do that now at quite high speed."
Take the simple example of a spark plug. For an engine to function properly, the spark plug gap must be set precisely. However, for a human inspector to have to use a feeler-gauge to measure the gap of each plug as it comes off the production line would be slow, error-prone, and expensive. MV can automate the process, speeding it up tremendously.
Manufacturers are looking to MV to reduce the risks with product recalls -- not only to eliminate defects but also prevent mislabeling of products that can damage a company's brand, cause regulatory violations, and introduce legal liabilities. This is one reason the pharmaceutical industry became an early adopter of machine vision.
Machine vision is helping manufacturers improve quality control. Catching bad parts early in the production process keeps them from getting incorporated into assemblies that might later have to be scrapped, resulting in great losses and poor yield.
For example, TRW Automotive, based in Livonia, Mich., employs machine vision to automate quality checks on its assembly line for braking systems. A combined system of cameras, lighting, processing systems, and software allows rapid checking of the positioning of components on the assembly line, the angle and length of the components, and the quality of coatings. Along the way, the vision system repeatedly checks barcode labels to keep track of parts and assembles and ensure traceability.
Richa Gupta, analyst at Natick, Mass.-based VDC Research, wrote that machine vision is becoming an important area of investment for manufacturers trying to achieve consistency on the shop floor. The adoption of MV, Gupta noted, is driven by a need for manufacturers to get better at error-proofing their production operations and to improve quality control and assurance.
One of the implementation challenges for manufacturers, Gupta noted, is the proliferation, complexity, and fragmentation of MV offerings, with "diverse business process applications, standards, and connectivity options, among which users must decide." Strategic decisions should be based on a good understanding of "how machine vision and its enabling applications can reduce pain points, increase automation, and lift profits," the research pointed out.
Most purchasers will need the help of a solutions integrator, Gupta said. MV is highly complex, spilling out into domains as diverse as optics and lighting, image processing, and network connectivity, not to mention the larger-scale challenge of integrating MV into the overall company automation infrastructure.
Given such challenges, the design, development, and deployment of a solution requires specialized knowledge and skills that would be hard for most companies to develop internally. Selection of a solutions provider or integrator should rest on "their application or vertical market expertise, product offerings, professional services capabilities, and the ability to undertake MV deployment and integration with minimal disruption to production operations," according to Gupta.