Focused standards needed

In the late 1970s, computers could perform limited image-processing tasks. Researchers, therefore, developed a number of algorithms for image analysis and interpretation.

Jun 1st, 2001
Th Andywilson

Rather than relying on established consortiums to dictate standards, the image-processing and machine-vision community should be developing its own standards.

Click here to enlarge image

In the late 1970s, computers could perform limited image-processing tasks. Researchers, therefore, developed a number of algorithms for image analysis and interpretation. However, after the advent of open-bus computing, system developers had to integrate off-the-shelf cameras, developed primarily for the closed-circuit and video markets, with Multibus or VME-based imaging boards to build rudimentary machine-vision systems.

Even then, the lack of image-processing standards forced vision companies to adapt whatever products were readily available—such as broadcast cameras—to their needs. As a result, the standards that were adopted were based on NTSC, PAL, and SECAM specifications—the three main television broadcast standards that are mutually incompatible.

Impact of the digital world
In today's digital world, the number of standards continues to increase. Systems integrators can choose from various digital standards, such as FireWire and Camera Link, or a range of analog standards such as RGB, RS-170, or other broadcast standards. Of course, each offers its own benefits and limitations.

In addition, the digital world has brought forward several standards that are supposed to ease the burden of systems integration. Last month I reported on several standards designed to ease the integration of machine-vision systems on the factory floor (see Vision Systems Design, May 2001, p. 80). These include Modbus, Profibus, DeviceNet, and ControlNet. And, while numerous vendors are working to develop more standardized Ethernet-like buses, much of the legacy vision and imaging equipment currently in use becomes incompatible. This installed equipment mandates the use of bridges to transfer information from one system to another.

Some computer manufacturers are proposing new standards, such as PCI-X and Infiniband, to eclipse current PCI-, ISA-, or EISA-based systems. Consequently, in all areas of digital development, multiple standards exist for host buses, control engineering networking, computer systems, and image storage. And, as the ease of use of image-processing systems increases, their requirements will more likely as not be met by more standard methods to interface peripherals to host computers.

Sometimes, a standard targeted for one market ends up being used in another. This was certainly the case with broadcast-based cameras being adopted by the machine-vision industry. Moreover, the IEEE-1394 interface standard now being promoted for camera interfacing was initially aimed as a replacement high-performance disk interface.

Designers of machine-vision systems are now confronted with many standards and their trade-offs and how to best integrate them with deployed legacy systems. Not only must designers be aware of computer, broadcast, machine-vision, and networking standards, but they must also have a firm knowledge of how to integrate such systems with other automation systems. Standards are necessary, but in some high-technology industries perhaps fewer standards are needed to address the same problems.

Rather than relying on established consortiums of computer and networking companies to dictate these standards, the image-processing and machine-vision community should be developing its own standards. The advent of the heralded Camera Link standard is a step in the right direction. Camera and frame-grabber vendors joined forces to develop a specific standard to ease the deployment of machine-vision systems.

Image and vision manufacturers should again similarly unite under the umbrella of the Automated Imaging Association or the Society of Manufacturing Engineers to promote further standards development in image transmission, networking, and display. Otherwise, systems designers will be challenged by standards that are not well suited to the vision/imaging market.

by Andy Wilson
EDITOR
andyw@pennwell.com

More in Scientific & Industrial Research