Machine-vision industry bounces back

An exclusive interview with a technical leader.

Larry Maloney, Contributing Editor -- Test & Measurement World, 9/1/2011 12:00:00 AM

Laval Tremblay VP of Engineering Matrox Imaging Dorval, QC, Canada Laval Tremblay received his electrical engineering degree from Montreal’s École Polytechnique in 1980 and joined Matrox Imaging in 1983 after working at Mitel Semiconductor and Northern Telecom. By 1986, Tremblay had been promoted to project leader, working on the company’s first PC-based image-processor add-on hardware. By the time he was named director of engineering in 1994, Tremblay had already begun to set Matrox on a course to redefine industry standards in image processing. He was named VP of engineering in 2004. Contributing editor Larry Maloney conducted an e-mail interview with Laval Tremblay on trends in machine-vision applications. This version of the article includes questions and answers that did not appear in print.

Q: Has the machine-vision industry recovered from the global recession?
A: The industry has recovered exceptionally well as a result of the manufacturing sector leading the overall recovery. Major players in the industry—including Matrox Imaging—are experiencing record sales. Manufacturers, especially those in the electronics and semiconductor sectors, better managed their inventories, having learned from the dot-com bubble. This discipline has resulted in an upsurge in demand for machine-vision components. Moreover, although the latest recession severely affected all economic activity, it had a greater negative impact on the financial, real-estate, and construction sectors. The question to ask, though, is how long can this strong resurgence be sustained?

Q: Which industries are leading the way in adopting inspection solutions based on machine vision?
A: Industries like semiconductors, electronics, and flat-panel displays continue to be big consumers of machine-vision inspection technology. Industries that are tightly regulated and those that need to mitigate potential liability—like food and beverage and pharmaceuticals—are also increasingly using vision. Plus, we’re seeing a rebound in automotive demand with the growth in electronics content for battery management, start/stop systems, onboard infotainment, and communications. LED lighting and new power-train investments, which promote smarter use of our scarce energy resources, are also creating new demand for machine vision in the automotive sector.

Q: What are some typical applications for the Matrox Supersight system?
A: The Matrox Supersight computing platform is designed for high-throughput and computationally demanding vision applications, such as semiconductor wafer and mask inspection and flat-panel-display inspection. Among its key features is a unique PCIe Gen 2 switched-fabric backplane that removes I/O bottlenecks between installed devices. Supersight is also well-suited to medical imaging, where applications demand high data and data/task-level parallelism and employ multiple CPU cores or GPU accelerator boards, or a mix of both of these processing technologies.

Q: Can you briefly sort out the landscape of vision interface standards?
A: For moderate-bandwidth applications, GigE Vision has become the preferred choice over IEEE 1394 and USB because of cabling advantages. GigE provides enough bandwidth for all but the highest-speed cameras, and cables can be as long as 100 m. Camera Link is holding its own for more exacting applications because of the better determinism, despite the fact that it requires specialized computer-interface hardware.

For high bandwidth, Camera Link is still king, although it is losing some of its dominance. This situation has prompted the creation of the CoaXPress and Camera Link HS standards, which offer greater bandwidth over longer distances. Initially, we see CoaXPress as being complementary to Camera Link, but we do see it eroding the Camera Link market as time goes on. CoaXPress offers several advantages, such as the ability to transfer higher data rates over longer distances with simplified cabling.

Q: How have smart cameras changed the machine-vision field?
A: It is not the camera hardware that is changing machine vision, but rather the software environment that is associated with the camera. Customers have certain expectations, such as quick-and-easy application development, which includes algorithms, the human-machine interface, and connections to other equipment. Software plays a major role in meeting those needs.

The following questions and answers did not appear in the September 2011 issue:

Q: Where are frame grabbers still prevalent?
A: High-bandwidth interfaces like CoaXPress and Camera Link HS will still require frame grabbers. Frame grabbers are also needed in such vision applications as semiconductor and flat-panel inspection, which require high resolution and high-data-rate cameras.

Q: How is Matrox responding to the demand for machine-vision solutions that can handle high bandwidth?
A: Historically, Matrox has provided a board-level solution. We’ve since realized however, that a more system-level approach is required. Our box-level solution offers a level of electrical power that far exceeds the power available from a single-board solution. This approach is needed to deliver the computing power required to keep up with the influx of data in demanding machine-vision applications.

Q: How prevalent are vision applications that require 3-D vision or color capability for inspection?
A: Although such applications are gaining ground, 2-D monochrome still reigns supreme in machine vision. Machining and electronics manufacturing are driving 3-D vision, while packaging applications are creating more demand for color vision systems.

Q: Can you briefly describe a leading-edge application featuring Matrox smart cameras?
A: A good example involves ThyssenKrupp System Engineering, Auburn Hills, MI, which was asked by a manufacturer of construction and heavy equipment to build an engine assembly plant. Besides its role as a line builder, ThyssenKrupp had the mandate to provide this factory with a machine-vision solution. The addition of vision came about when the customer expressed concern with the final test machines, which were failing engines due to operator errors during assembly. ThyssenKrupp presented a camera-based solution that would use vision inspection at the station where the parts were being installed. Any errors would then be fixed at the assembly station.

The first steps of the vision solution were developed using six Matrox Iris GT smart cameras and Matrox Design Assistant, an integrated development environment that is bundled with the smart camera. Design Assistant lets users create machine-vision applications by constructing a flow chart instead of coding programs or scripts using a language like C++ or C#. As an engine moves down the line, it goes through multiple smart-camera-based inspection stations. These applications vary from absence/presence to gauging/measurement.

ThyssenKrupp evaluated other smart cameras for the vision system, but the company felt that the Matrox Iris GT offered some key advantages, including the fact that it has a lot more memory than other smart cameras. This lets operators load multiple jobs (inspections) in memory simultaneously. Other smart cameras’ memory limitations would have required the operators to load one job at a time. In addition, with many other cameras, images would have to be stored on a separate PC, since there would be inadequate on-camera storage. Design Assistant’s flow-chart logic also allows users to branch and execute different inspections within an application. In contrast, other smart cameras only support the development of stand-alone inspections. As a result, an application needs to be shut down before the next one can be loaded and executed.

ThyssenKrupp’s engine-line vision system was deployed in September 2010. Deployment went smoothly, and application troubleshooting was handled remotely. And so far, there have been no technical issues that have required ThyssenKrupp to send staff on-site to troubleshoot.

Q: The Matrox Imaging Library was a 2010 finalist for Test & Measurement World’s  Test of Time award. What makes this tool so valuable for users?
A: MIL, which was introduced in 1993, eliminates the need for users to develop algorithms from the “ground up.” It also is platform-independent and offers a very intuitive application programming interface. Once a user learns one tool, it’s very easy to figure out how another tool works. And last but not least, MIL is a robust, field-proven tool for vision applications.

Q: Besides tools like MIL, which addresses ease-of-use concerns, what further steps must your industry take to encourage greater use of machine-vision technology?
A: The industry must develop strong relationships with academia to ensure that the technicians, engineers, and scientists of tomorrow are familiar with vision technology. Industry associations must continue to offer courses, certify industry experts, and organize trade shows and conferences. Finally, manufacturers, such as Matrox, need to promote vision by publicizing our success stories and offering our own training programs and workshops. T&MW