This article explores benefits of vision system and role it can play in manufacturing industry and beyond. Vision systems rely on cameras and image processing software with interfaces to perform manufacturing tasks. The article also highlights that manufacturing insiders see an ever-expanding number of uses for vision technology in the not-too-distant future. Vision systems can be programmed to perform narrowly defined tasks such as to count objects on a conveyor or to search for defects in a product or in packaging. The vision systems, through software joined to other equipment and devices, can be used to help make decisions about—and act upon—what they see. Machine vision systems are also making their way into biomedical applications. Progress in biomedical research has been highly dependent on engineering techniques and these types of vision systems have immense potential to return unprecedented information for cellular biology.
Magicians like to claim that the hand is quicker than the eye. But for Manner AG, the maker of wafers, biscuits, and cookies based in Perg, Austria, it was the assembly line that was too fast to follow.
The company’s plant is capable of maintaining packaging speeds of up to 400 packages of Manner wafers per minute. The wafer itself is packaged in many different package varieties.
“If the conveyor belt is operating at a speed of, for example, 270 packages per minute, then just a few seconds of defective production means several dozen rejected packages,” said Reinhard Gassner, Manner plant manager. “This is not just a question of cost; it also has a negative effect on the production flow.”
Like many manufacturers, Manner recently implemented a new vision sensor system intended to enable many different package characteristics to be checked simultaneously and quickly. In fact, over the past five years or so, machine vision technologies have stepped up to play a key role across a number of industries.
The systems—once found mainly on automotive and industrial assembly lines—are moving beyond those traditional implementations. In recent years they’ve made their debut for use in quality and appearance control across a range of industries, said Bob Rongo, the owner of Decision Technology LLC, in Indianapolis.
Improvements in the cost and capability of the machine vision technology has helped spur wider adoption, to be sure. But another reason is that the technology that drives the systems is becoming more accessible to manufacturing engineers without a programming background, said Kyle Voosen, business development manager for vision at National Instruments of Austin, Texas. The company makes software that drives the systems.
As vision systems get easier to program and to use, manufacturers are finding new and interesting ways to incorporate them into their manufacturing environments, Voosen said. Also relatively new is the capability for manufacturing engineers to create and customize systems to their own needs rather than buy a generic system suited to their industry, Voosen added.
And the systems are moving beyond traditional manufacturing. They also have a role to play in biomedical applications, said Masaaki Yoshigi. He’s an assistant professor of biomedical engineering at the University of Utah in Salt Lake City
Manufacturing insiders see an ever-expanding number of uses for vision technology in the not-too-distant future. Many more companies than in the past are creating turnkey vision systems meant to operate within a very specific manufacturing sector, Rongo said. His company, Decision Technology, for instance, makes systems specific to the wholesale bakery industry.
“There’s growth potential in baking because machine vision can be used for overall quality control and for verifying the final product looks right,” Rongo said. “But it also can be used within the baking process itself and for packaging. It can really be customized to fit a number of needs.” A vision system may be thought of almost as an inspector or operator at work on the production or packaging line. As the name implies, vision systems have the capability to see what’s happening on the line, albeit in a limited way far removed from how a human relies on vision.
These systems instead call upon digital cameras, smart cameras, and image processing software to carry out their duties. They can be programmed to perform narrowly defined tasks such as to count objects on a conveyor or to search for defects in a product or in packaging.
But machine-vision systems go beyond merely noting defects. Through software tied to other equipment and devices, the systems can be said to help make decisions about—and act upon—what they see. If a system is programmed to ensure each loaf of bread is of uniform color, for example, and it detects a loaf that is too light or too dark, it relays that information to a robot or an operator further down the line. The operator or robot then knows to remove that loaf from the line, Rongo said.
The economy has also played a role in the way manufacturers are now using vision systems. Companies now are looking beyond inspection capabilities to find ways the systems can be used to cut manufacturing costs, National Instruments’ Voosen said.
“Now it’s about things like how vision systems can save money by cutting down on scrap,” he said.
His company sells the components and offers the software training that enables users to build their own vision systems suited their unique needs. This capability allows the systems to be fme-tuned to a range of applications. The software is easy to program for particular needs, no matter what the user’s technology background or comfort level may be, Voosen added.
“If you can already program in C-plus you wouldn’t care about these advances,” he said.
But it goes without saying that most manufacturing engineers or other employees looking to set up unique vision systems don’t have the extensive programming background needed to fine-tune or to create their systems.
“But now whatever a manufacturer’s use—pattern matching, reading bar codes, edge finding, or measuring distance—the algorithms can be programmed by lay people to the manufacturer’s needs,” Voosen said.
The software works like this: Say a manufacturer needs a system that locates crooked labels affixed to parts. The engineer will load images of parts into the software and then look for algorithms—again within the software— that properly identify the misplaced labels.
“Once you find an algorithm you like—like one that checks angles or colors—then you try it for several images to make sure it works for all your images,” Voosen said.
The system can also be programmed ahead of time from a desktop computer contained within a work cubicle, much the way mechanical engineers design and test complex systems on their desktops with today’s CAD and analysis software.
Manufacturers of all stripes are now fme-tuning vision systems for their specialized needs.
Can it be Reused
Take Ricoh Industrie France of Wettolsheim, France, which makes digital photocopiers and toner cartridges for printers and fax machines. Ricoh uses a vision system to make sure parts can be suitably recycled, said a company engineering spokesperson.
One complication in recycling these parts is the cleanup strip affixed to the toner cartridge via a metal support.
“We realized that some cartridges had common defects affecting thé strip. The strip’s position changes during normal use and wear, which manifests as notches on the strip and affects print quality by leaving marks on the paper,” said the engineering spokesperson.
The company already used a vision system to check for contact between the strip and the machine that reloaded the toner cartridge. Now it sought a way to isolate the cartridges that could be recycled from those that were no longer usable.
Engineers set out to design a vision system that could check for notches on the strip as they came down the line. Those with notches wouldn’t be recycled.
Ricoh worked with Alliance Vision of Montélimar, France, which supplied the video cameras, lighting, and image-acquisition boards necessary to create an overall vision system. Ricoh also employed the LabView and IMAQ Vision software packages from National Instruments.
Now as the cartridges come down the line, an analog video camera connected to an image-acquisition board essentially takes a picture of each cartridge for the line operator to study. Vision analysis software tied to the system measures points at the ends of the strip and at its center. These measurements—along with visual inspection of the image—tell the operator whether the cartridge is compliant and can be recycled, according to the spokesperson.
“Various tests led us to define a maximum depth of a strip’s edge notches. If the strip is compliant, we can recycle the cartridge. If not, we dismount some of the reusable unit components for retrieval,” the engineer said.
The new system enables the strips to be analyzed and either accepted or rejected—all in less than a minute, he added.
In a bid to keep things simple for wholesale bakers who use their systems, Cognex Corp. of Natick, Mass., has integrated its vision hardware and software into one package. Bakers can program the system for their own needs—whether quality control or packaging.
“Our systems can be made to display a picture or a message to an operator, but they can also be more complex,” said Lisa Eichler, director of vision systems marketing at Cognex. “They can talk to the controller on the line that controls all the equipment or they can talk to a robot to tell it where the bag of chips is that it needs to put in a carton. Or they can be talking to a software package that controls many different type of devices.”
Cognex routinely provides vision systems for a variety of baking applications. One of its customers is Manner AG, which produces 8,000 tons of flat wafers and biscuits each year in its large baking oven.
The company recently implemented a vision sensor called the In-Sight 5400 and combined it with PatMax pattern-matching technology on its packaging line. The hardware and software are from Cognex.
The vision system is programmed to recognize each package, no matter where it sits on the line, and to detect any fault in the packaging itself, Gassner, the Manner plant manager, said. The pattern-matching software pairs the actual packaging the vision system “sees” against its knowledge of how the package should look. In this way, it recognizes appearance defects.
With the system in place, individual packages of wafers flow through the inspection station on the conveyor belt without needing to be fixed in a particular position.
The system can determine that the packaging properly wraps the product and is centered correctly. It can also determine the presence of the picture of the hazelnut on the hazelnut wafer, for example, thus ensuring packages correspond to the wafer type inside. The system also checks packages for dents and tears.
When the system alerts Manner employees to a dented or otherwise unacceptable package, they have to react quickly. A damaged package may become jammed further down the production line, resulting in additional product losses, Gassner said.
Since implementing the vision system, productivity on the Manner line has increased by about 5 percent because interruptions in product flow—to manually check for crushed packages, for example—is kept to a minimum, he added.
Machine vision systems are also making their way into biomedical applications, said Yoshigi, the University of Utah professor.
His department has constructed a vision inspection system to assess the alignment and arrangement of cytoskel-etal fibers in a cell. The fibers are filamentous proteins found in any kind of cell. Remodeling and rearrangement of cytoskeletal fibers is a major topic in molecular and cellular biology.
But in order to rearrange these fibers, biomedical engineer must first inspect their orientation, Yoshigi said. That’s where the vision system comes in.
Engineers in his department first grew vascular smooth muscle cells on a glass cover slip using standard cell culture techniques. They then stained cytoskeletal fibers of the vascular smooth muscle cells using fluorescent antibodies. .
The researchers used cameras inserted into microscopes to capture and store images of the stained fibers. Software from National Instruments was used to process those images to ensure they depict exactly the area of the cell the researchers need to see.
“We can select the cell to analyze and then click around the region of interest,” Yoshigi said.
The fiber orientation analysis vision system automati-L cally eliminates unnecessary image areas and calculates the gravity center of the region of interest, he added.
“The brightness of the original image can vary depending on the antibody reactions or the selection of fluorescent dyes,” he said. “With vision inspection we can select several brightness adjustments.”
Progress in recent biomedical research has been highly dependent on engineering techniques and these types of vision systems have great potential to return unprecedented information for cellular biology, Yoshigi said.
As these systems become easier to use they’ll become ever-more prevalent in biomedical labs, in bakeries, on manufacturing lines, and perhaps even in manufacturing and research environments that haven’t yet thought to incorporate machine vision, Voosen said.