by Phil Heil and Richard Daigle

The economic axiom that what goes up must come down has held true in recent years. Money has been tight in many industries, but forecasters tell us that what’s been down will start to rise in 2004. This is good news for most businesses, but it’s great news for the machine vision industry.

In addition to the economic upswing, another factor contributing to what could be the best year ever for machine vision systems are smart sensors, or “smart cameras.” They’ve come into their own, offering far greater power for far less money than was the case even two or three years ago. As manufacturers prosper, they will invest more in quality control, inspection and robotic assembly--all areas in which smart cameras can make a significant difference, providing 100-percent inspection and no-fail robotic guidance.

This may sound like an optimistic fantasy, depending on your company’s situation, but it’s not. The National Association of Manufacturers forecasts a 6 percent production boost that will generate some 250,000 jobs in 2004, partially offsetting heavy job losses suffered in recent years. The Manufacturers Alliance/

MAPI reported its business index rose to 77 during the last quarter of 2003, the highest level since the index began in 1972. Short- and long-term futures for machine vision are equally bright.

New kid makes good

There was a time during the early 1990s when smart cameras had a reputation, perhaps deservedly, as expensive high-tech gadgetry that didn’t deliver in terms of performance. Just five years ago, the average cost of a machine vision smart camera system was out of many companies’ price range. Today, an even more powerful

system can cost less than $2,000. During a recent training luncheon at DVT Corp. in Atlanta, one student, commenting on a machine vision system’s capabilities, said, “It’s not a matter of knowing what application the camera can handle but what application it can’t handle.”

A proven track record for smart camera machine vision will also carry weight in the minds of decision makers, especially for applications in which meeting strict government standards--such as 21 CFR Part 11 or other quality control measures--is critical.

Machine vision systems are used in a variety of applications, including:

n Pharmaceutical packaging inspection. The machine vision systems are typically 21 CFR Part 11-compliant.

Automotive light bulb inspection that meets Six Sigma standards

Inspection of fill-levels of opaque bottles using smart cameras in concert with X-ray technology

Inspection to ensure Teflon coating is properly applied to medical inhalant canisters

Although machine vision applications have proved increasingly useful in recent years, the technology offers untapped potential in the manufacturing industry worldwide. As assembly line speeds have increased, human inspectors can’t keep pace with a 100- percent inspection rate.

Simple sensors and other rudimentary machine vision tools offer a basic level of inspection, but as applications grow in sophistication, smart cameras are proving their worth on the assembly line. Packaging, pharmaceuticals, automotive, electronics and plastics are all growing markets for machine vision smart camera systems. Lower integration costs and better after-sale support give the technology a distinct advantage over frame-grabber systems in many applications.

A smart camera or “smart image sensor” is much more than just a camera. It’s a digital camera, small enough to fit in the palm of your hand but also packed with the high-tech capabilities of an onboard computer processor and a CCD that captures digital images at the smallest unit: a pixel.

High-end smart cameras can measure down to a tenth of a pixel with total reliability and even smaller with varying degrees of reliability. Smart cameras can take precise measurements, sort by miniscule color variations and object shape, and perform optical character recognition, bar code recognition and other tasks--all at speeds faster than most assembly line inspection rates. Smart cameras can handle speeds up to 2,000 parts per minute, depending on the application.

Sales of smart image sensors for manufacturing applications have grown from 30 to 50 percent annually since their introduction more than a decade ago. Vision Systems International conducts annual financial analysis reports that review productivity, profitability and other standards for public machine vision companies. Its 2002 report, which looked at 31 companies that receive 5 percent or more of their revenue from machine vision/imaging, indicates a slowdown of growth as compared to 2001 numbers but, nonetheless, a 19 percent median gross margin. Despite absorbing blows from the sluggish U.S. economy, the future of smart cameras remains bright.

Multitasking sensors

The technology’s versatility and ease of use continue to drive its success in the marketplace. Sensors are used throughout manufacturing to provide feedback for process control and monitoring. The sensors can direct their information to programmable logic controllers through a hard-wired DC connection. In the simplest case, the sensor provides a yes/no indication.

For example, a proximity sensor will detect if a part is present or not. More advanced sensors can provide a variable result. For example, a fill-level sensor can indicate the fluid height inside a tank. Another common example is a bar code reader that collects a code and outputs it via serial to a controller.

A smart camera is similar to standard sensors in that it visually measures a quantity or reads information from a part. The difference is that a smart sensor can be configured to perform a variety of tasks simultaneously and communicate this information in nearly any format, as illustrated on page 21.

Smart cameras are used in a number of fields, including automotive, pharmaceutical, plastics, electronics and food/beverage. They’re proficient in detecting defects, taking precision measurements, and picking, placing and performing other positioning applications involving robots, OCR, data collection and 2-D data matrix. Currently, packaging is the fastest growing industry for smart camera machine inspection.

For the first time, smart cameras are being used in an X-ray unit that features affordable nondestructive testing for nontransparent applications. The combination of smart sensor technology and X-ray capabilities allows smart cameras to perform varied functions on items that are hidden to the naked eye.

The KMV Technologies InnerVision X-ray system, using DVT cameras, is an example of the marriage between X-ray vision and machine vision. The KMV unit can be used for most nontransparent applications, including:

Checking wire connector integrity

Checking fill-levels in nontransparent containers

Checking proper fill of injection-molded parts

Checking electrical connector pin alignment

Unlike smart cameras, traditional sensors can’t be programmed. Instead, they’re limited to performing one task at a time, making a single pass/fail decision based on adjustments to sensitivity. A typical application might read the bar code on a part, perform a measurement and relay this information to the factory network. The PLC would set the part aside for rework if it didn’t meet the specification, and the work cell controller would track the part based on the bar code. With a traditional sensor, this application would be done using contact probes and a bar code scanner connected to a PC--three different hardware units performing separate tasks with separate information channels.

A smart camera can perform these tasks simultaneously with the same set of tools, completing the application at a lower cost and providing more options for flexible manufacturing. It can also convey information from more than one inspection in one communication channel.

Two major classes of vision systems compete in the market today. Vision and photoeye systems can be taught using a single button or a video game controller, whereas smart cameras are connected to a PC. The former class is easy to set up but less flexible and limited in communications options. Such sensors are best used for solving simple problems when a great deal of automation hardware is already present to handle variations and communications, as illustrated at the top of page 22.

Integrating with larger platforms

Traditional sensors using serial connections make pass/fail decisions that can’t be integrated into one system but must instead travel to different systems on different paths. Because they’re not programmable, traditional sensors can’t make intelligent decisions about information. They simply make one decision based on input from one source. When a customer uses a PC and takes advantage of the familiar Windows interface, setting up a more advanced class of smart camera is easy. In this case, a powerful communications system’s flexibility can be fully configured. When standardizing on a common sensor platform, a single software package can be learned, and many different applications can be solved quickly with a common interface, as illustrated on page 22.

Smart image cameras handle information flow much more efficiently than traditional sensors. Smart cameras can perform more than one inspection at a time, sending that data in a single stream to PLCs or other devices. The data can then be easily stored for analysis or use with reject mechanisms or other components on the line. Smart cameras are commonly found in the automotive industry, where there’s an urgent need for quality, low cost and ease of use. And now, with the expanded capabilities of recent models, packaging and pharmaceutical industries also make use of the technology.

The applications vary widely, but some common problems solved by smart cameras include checking for a part’s presence or a defect’s absence. When first developed, the smart camera was designed to replace arrays of photoeyes that were used to verify that a part was assembled properly. The photoeyes did a great job until it was time to switch the line to a new part.

A smart camera can draw on hundreds of configuration files to determine what inspection is to take place. As tools were added to the design, tasks such as dimension checking and tolerance verification were accomplished with low-cost platforms. Formerly done by humans or analog frame grabber-based vision systems, the smart camera proved that direct-image acquisition provided a level of consistency that hadn’t been seen in a hardware device costing less than $10,000. This led to further improvements to handle bar code reading and optical character recognition in a single package. Applications that would normally be done with four or five different devices could be solved with one.

Throughout the 1990s, the Ethernet market outgrew its humble beginnings of connecting PCs in schools and offices. Industrial devices emerged that could communicate over this now ubiquitous interface. In late 1998, DVT introduced the first vision smart sensor with onboard Ethernet. The smart sensor had outgrown its role as a simple yes/no decision maker into an information provider for the entire factory.

In the coming years, as industrial quality standards merge, the Ethernet may well be the hardware interface that binds all the equipment. Smart sensors will evolve along with the standards to support communication, from the least expensive PLC to the most costly enterprise resource planning software package. As software and hardware are added, these sensors will be able to handle increasingly complex visual inspection tasks. At the same time, the falling costs of electronics and the availability of free training will allow more of the sensor market to be overtaken by inexpensive smart sensors.

The success of new technology is usually tied to the ebb and flow of economic fortunes. When money flows, technology grows. As the machine vision industry is fueled by new capital expenditures, smart camera technology will continue to provide more robust solutions for the manufacturing world.

About the authors

Phil Heil earned his mechanical engineering bachelor’s degree at Carnegie Mellon University and his doctorate in mechanical engineering at Georgia Tech. He has worked at DVT Corp. in the Applied Engineering Department since 1996. Heil has since worked in training, technical support and application support for DVT.

Richard Daigle has a bachelor’s degree in journalism from Southeastern Louisiana University and a master’s degree in communications from Georgia State University. After serving as reporter and editor for a daily newspaper in South Georgia, he worked in public relations for the State of Georgia Department of Natural Resources, Coastal Resources Division. Richard was formerly a writer for a multimedia company in Atlanta. Since joining DVT, Daigle has launched the company’s quarterly magazine, Vision in Action.

Portions of this article © 2003, Society of Manufacturing Engineers.