Vision Systems

Although machine vision systems long have been a common fixture in electronics and semiconductor industries, dropping prices of components and integrated systems as well as improved performance are making them even more attractive additions to plants in a wide variety of industries, including automotive, consumer packaged goods, food and beverage, pulp and paper, and pharmaceuticals.

By Dan Sussman, for Control Engineering August 15, 2005
Sidebars:
Vision systems checklist

Although machine vision systems long have been a common fixture in electronics and semiconductor industries, dropping prices of components and integrated systems as well as improved performance are making them even more attractive additions to plants in a wide variety of industries, including automotive, consumer packaged goods, food and beverage, pulp and paper, and pharmaceuticals.

In a typical application, a video camera placed above or to the side of an inspection point captures an image of a part and sends it to a machine vision processor. Generally, this processor is either an embedded task-specific device or a PC equipped with vision-processing software. A combination of hardware and machine vision software analyzes the image to perform one or more tasks, including determining the presence or absence of parts on a line; pattern matching to ensure that parts are properly aligned; color inspections; barcode and character reading; gauging; and detection of a variety of defects.

The configurations of vision systems are as numerous as the plants, mills and fabs that use them. Prices range from about $2,000 for entry-level packages, consisting of a single camera, light, and basic processing unit, to about $15,000 or more for systems consisting of multiple, high-performance cameras, multiple lights, sophisticated machine vision software packages, and networking capabilities.

When shopping for a machine vision system, users should consider a number of factors before making a purchase. These include:

Camera selection —All cameras are not created equal. Numerous low-cost analog cameras are on the market, but for many users, capabilities will fall far short of what is needed. Many applications require fast frame speeds and high resolution. Ensure the camera can meet specific application requirements.

Ease of use —A number of vendors point to ease of use as the most important competitive differentiator in the marketplace. In evaluating machine vision systems, buyers should consider how quickly systems can be set up and how easily they are maintained. The more complex a system, the more likely it is that the user will have to pay a system integrator or consultant to implement it.

Software: configuration versus programming —Configurable software generally features pre-built functions that simply can be selected and fine-tuned using point-and-click menus. Programmable software enables users to write custom code in high-level languages such as C, Visual Basic, or proprietary vendor-specific languages. Configurable applications are easy to use, but not as powerful or flexible as programmable software. However, the latter is harder to use and requires services of experienced programmers.

System expandability —While a plant might now need only a simple one-camera system to detect part presence or absence, buyers should carefully consider whether they might eventually want to implement applications that require multiple cameras, additional lighting, and more powerful software. If expansion is a possibility, buyers should consider spending more for systems that can accommodate multiple cameras, powerful application software, networking, etc.

“On the tools side, you also want to make sure that you’re not right at the edge of performance,” says George Blackwell, marketing director with Cognex, a vision systems manufacturer. “Make sure that your vendor can supply a large library of tools and that they have enough flexibility for a wide variety of applications.”

Vendor stability —Numerous vision system vendors have been acquired or have disappeared during the past few years, raising questions about upgrades for or migration from their systems. When purchasing vision systems, buyers should consider the long-term stability of the vendor and the size of its existing user base. If possible, they also should consider the vendor’s support capabilities.

Type of environment in which the vision system will be used also should be considered in the buying decision. For example, in semiconductor applications, fanless computers or embedded processors are a must to minimize possibility of particulate contamination; in food and beverage plants, where clean-in-place washdown is required, vision systems must be waterproof; and in pharmaceutical operations, where the FDA requires full product and process traceability, vision systems may be required to function on networks and exchange data with other plant computing platforms, such as plant historians.

Machine vision systems allow humans to pursue more suitable roles beyond inspection, and, in most cases, vision systems do a better job of inspecting parts and products, especially at high speeds. Still, outside forces affect adoption in plants and mills. Long-term payback from vision systems can be substantial, but investments compete with other priorities in the plant and can wind up delayed, particularly in economic crunches, says Steve Geraghty, director of vision systems manufacturer ipd, part of Dalsa Coreco.

Another factor that affects adoption of vision systems is customer dissatisfaction with off-spec, poor quality products. In those cases, machine vision systems can quickly rise to the top of a plant’s list of investment priorities.

Vision systems checklist

Camera quality, price, and applicability

Ease of use, implementation

Software: programmable vs. configurable

Expandability

Vendor stability

Networking capabilities

Environmental considerations (enclosures for washdown; ability to withstand heat, vibration; fanless operation, etc.)


Related Resources