They can tell if bottles have sealed caps and accurately printed labels. They can detect smudges and differentiate items by color. They are a key part of sortation in processing lines.
Machine vision systems are becoming more popular in quality-control applications. “The machine vision industry has witnessed steady growth in the past three years, mainly due to the reduction in the cost of computing and electronic components and increase in capabilities of vision systems,” according to a September 2007 report entitled “Advances in Machine Vision Systems” authored by Vishnu Sivadevan, research analyst for the global growth consulting company Frost & Sullivan (San Antonio).
To See…Then Act
Although barcode and RFID scanners are sometimes described as ‘seeing’ devices, experts say they are fundamentally different from vision systems.
Vision systems use optical devices, such as cameras, to collect data and interact with hardware- and software-based control systems. The control systems ultimately direct product, via conveyor lines, throughout a facility.
John Mosher, solution development manager at Holjeron (Tualatin, Ore.), says the main difference between scanners and vision systems is intelligent decision making. “Scanners look for written material, while vision systems look at the product itself,” he says. “Vision systems look for product flaws. And, while RFID tags can hold a lot of information, they can’t ‘see’ if there is a flaw.”
In the food processing industry, for example, vision systems can detect the color of fruits and vegetables and transmit a message to a controller that will then divert or reject those items that don’t match the color preset into the controller. Items can then be diverted to a rework area or
SICK’s new IVC-2D reader can ‘see’ product irregularities and read barcodes and date codes, all at one inspection point.
scrapped altogether. “Once a camera sends the information to the control computer, the system makes a decision on where the product goes,” explains Mosher. Holjeron specializes in the intelligent controls that transmit those decisions to diverters on conveyor lines.
“All vision systems are designed to give machines ‘the ability to see’ and initiate action according to the perceived images,” according to Sivadevan. Because of this ability to ‘see’ many different features of a product, regardless of item orientation— and then initiate an action based on the image ‘seen,’—vision sensors are finding their niche in manufacturing. “Machine vision has gained importance due to the norms of 100% inspection and zero defects by current manufacturing standards,” Sivadevan wrote.
Rob Beideman, vice president of automation solutions at SICK (Minneapolis), says parts inspection is a typical application for vision systems. “Misplaced labels can be identified,” he says. “Vision systems can gauge whether a seal is placed on a bottle.”
“In quality-control processes that were manual,” Beideman explains, “users are finding they can get more from the automated capture of images and the associated analysis or increase throughput, because once you capture an image, that data is stored, and further processing is possible.”
Looking forward, quality control may be only one application for vision systems, as end users “want to capture more information or make more efficient decisions,” Beideman says. In fact, he says, manufacturers can often discover other operational benefits along the way. “Say a company wants to understand why certain products are being damaged,” he continues. “They can capture images of packages as they transport them through the network.”
Omron’s new ZFX machine vision sensor features a touch-screen human-machine interface (HMI) for easy programming and operation.
Photoelectric Vs. Vision
“Vision technologies are subsets of photoelectric sensors,” explains Gary Frigyes, product manager of photoelectric sensors at Pepperl+Fuchs Inc. (Twinsburg, Ohio). “Photoelectric sensors have defined fields of use, while vision sensors can evaluate multiple features of products for sorting or qualification.”
Photoelectric sensors typically detect a single point of operation and have limited built-in intelligence, says Tom Kahn, vision systems marketing manager at Omron Electronics (Schaumburg, Ill.). Vision systems, on the other hand, can perform discretionary validation of shape, color and size. “It’s a simple yes/ no versus an intuitive functionality,” he explains. The two technologies often function in tandem, as photoelectric sensors typically trigger vision systems, says Kahn.
One version of the new VOS300-Series vision sensors from Pepperl+Fuchs— the VOS301—verifies a single feature of a target object, such as a straw glued to a drink box. Another version—the VOS310—checks five unique features of an item. “For example, in a drinkbox application, the VOS310 could be used to determine if the straw is glued onto the box, the tamper-proof seal is in place, the barcode is printed, the package is labeled correctly and whether or not the package is closed,” says the company in its product literature.
“Photoelectric systems, coupled with imaging technology, can provide a flexible solution and may be the most efficient way to solve a particular challenge,” says Beideman. SICK’s new IVC-2D reader can inspect products for irregularities as well as read barcodes and date codes, all at one inspection point.
And, the company says its new IVC- 3D-30 short-range, 3D camera is ideal for detailed inspection applications, such as small-parts compliance. Capable of measurements down to 15 microns, the camera can capture 5,000 profiles per second. SICK’s new IVC-3D-300 smart camera has a larger field of view for pallet inspection, dimensioning and counting.
Due to ongoing downsizing in manufacturing, makers of vision systems are being pressed to create high-tech equipment that is both intuitive and easy to use, according to Omron’s Kahn. “We are seeing the elevation of operators to technician-level capabilities,” says Kahn.
Frost & Sullivan research analyst Sivadevan appears to back this up in his report: “Machine vision systems now incorporate user-friendly features that minimize operator training.”
Pressure to shorten commissioning time by simplifying device operation led to the development of Omron’s new ZFX machine vision sensor. Hardware has been condensed from nine components to only two—a camera with “intelligent lighting” and a controller with a built-in human-machine interface (HMI)—also sometimes referred to as a graphical user interface (GUI). Whatever you call them, easy interfaces allow operators at any level to perform vital inspection procedures—even programming— with little or no technical expertise.
The touch-screen HMI allows operators to view a selection of color icons and choose from nine inspection tools: pattern search for shape and character inspection; area search for size inspection; position, width and count functions using edge inspection; color and brightness inspection; and defect inspection to detect smears, scratches, chipping and burrs on product.
According to Omron, its vision sensor can be used for product or package sorting, label inspection, high-speed bottle inspection, electronics inspection, random position reference and high-speed robotic tool guidance. In a beverage packaging operation, for example, the vision sensor can detect the presence of the cap as well as any labeling errors or ink smears on the bottle. Thanks to the vision sensor’s three-step setting procedure for lighting and measurement, “anyone can rapidly perform a high level of image processing,” reads Omron’s product literature.
Likewise, the new VOS300-Series vision sensors from Pepperl+Fuchs combine camera, illumination, digital outputs, process data and five evaluation methods into a single sensor housing, which results in “an out-of-the-box” system that is “easily configured without programming knowledge and operated without the need for detailed image processing experience,” according to the company.
As machine vision systems become more user friendly and more valuable to the manufacturing process, they are “becoming critical to the emerging requirements of industrial automation and robotics,” wrote Sivadevan. “With quality standards being amended for the better on a continuous basis, the implementation of machine vision systems is becoming inevitable.”