Is a Smart Camera Right for You? Washington DC

Smart cameras are taking on increasingly more demanding industrial tasks. By definition, they are self-contained units that perform image processing on board. The data can be sent on to a third-party device like a PC or, for example, when performing pass/fail exams on an assembly line, can send a shut-down order to the machine. Their size, robustness and—in many applications—ease of use, often can make them more attractive than a PC-based system, says Ben Dawson, Director of Strategic Development at DALSA IPD (Billerica, Mass.).

We talked to a number of camera manufacturers, who discussed the advantages, disadvantages and future of smart cameras.

ADVANTAGES

Smart cameras are gaining ground in the industrial arena, says Michael Schwaer, product manager at Basler Components (Ahrensburg, Germany). "They are used for quality control, object measurement and other demanding applications. They contain an integrated processing unit that allows them to evaluate specific algorithms with respect to captured images. So a smart camera is able to produce control instructions or a measurement result without the need for additional hardware. The concept is more flexible and cost-effective than a PC-based system solution.

"An additional advantage is reliability," he adds. "A system with several smart cameras is more reliable than a PC-based solution. If one camera fails, the rest of the cameras will still work properly. If a PC fails, the entire system will collapse."

Space-savings is another advantage, says Fabio Perelli, Product Manager at Matrox Imaging (Dorval, Quebec, Canada). "A smart camera can be considered a camera-sized computer, one that can acquire and process image data before it leaves the camera," he says. "Since a smart camera is a self-contained device, they provide an alternative for applications that are limited by physical space."

It also can save time, Perelli adds. "An automation or manufacturing engineer can set up a complete inspection system—one that uses complex image-processing algorithms—without writing a line of code. A smart camera, then, can shorten development time."

And, says Steve Kinney, Director of Technical Pre-Sales and Support at JAI (San Jose, Calif.), cost also is a consideration. "The basic advantages of smart cameras originates in a concept of not providing or paying for more than you need [for example, a PC often has more processing, I/O and networking, disk drives, memory, high-end processor etc…than is required for many basic machine vision tasks]. Most of the time, cost is a driving factor. While full-blown smart cameras are usually not cheap, the expectation is that they are cheaper than a solution that includes a PC, monitor, vision software or custom algorithm and camera. This is true in some cases, but with the constantly driven, consumer-based PC cost/performance ratio, and looking at total system/installation costs, it often is not true."

DISADVANTAGES

Smart cameras may not be flexible enough to meet a customer's long-term hardware requirements, Basler's Schwaer says. "A PC-based system may be better able to adapt to an increase in the hardware required for the processing task. For example, a PC-based solution can often be extended easily if there is a need for a more powerful processing unit."

In that vein, agrees Matrox's Perelli, a smart camera is limited by the power of its CPU and how much data must be collected. "As the required data increases, the processor's performance also increases," he explains. "This is why heat dissipation around the processor can be a problem." Intel, he says, is working on methods of reducing power consumption, which will alleviate the heat problem.

JAI's Kinney points to limited video formats, saying smart cameras offer only a few choices of images. "It's silly to think that you can cover even a portion of the machine vision market with a few imager choices integrated into a camera body." He also says performance and functionality must be carefully considered.

"Making the highest performance and highest performance-to-dollar ratio cameras requires a focused effort on the camera technology. Many smart cameras use only lower performance CMOS images. A vision system is only as good as its eyes. And, similar to performance, many machine vision applications require functionality within the camera to capture or produce the best image, including high-speed asynchronous reset, flat-field correction, look-up tables, A/D control, and a variety of flexible trigger modes. Most smart cameras fall short in this area."

APPLICATIONS

"The smart camera is well-suited to many applications as it is small, and can generate pass/fail results," says Perelli. "This can be a presence/absence detection, an OCV/OCR operation, or reading bar or matrix codes. On the other hand, web inspection, and/or line scan applications are likely to offer better performance with a traditional PC-based vision system because of the large volume of data that must be processed."

Another example of a good fit is solar panel inspection on a production line, says Schwaer. "During the production process, a smart camera acquires images of the solar wafers and detects any cracks in the wafers. An advanced algorithm is used to make each inspection and to evaluate the results witinn a few milliseconds. The decision to accept or reject a wafer is sent to the production system via the camera's digital I/O—no image is transmitted."

Another example: Three smart camera from Vision Components (Ettlingen, Germany) were used to test sintered (a method for making objects from powder, by heating the material until its particles adhere to each other) metal moldings, measuring 5mm each, in an automotive plant. The parts had to be examined for external measurements as well as true running, axial runout, inner diameter and the occurrence of ridges, injections and cracks. The required tolerance for the inner diameter and the overall axial run-out tolerance for the flanks and the front sides is 50 microns. One camera handles the true-running test and the other two determine the inner diameter and check both flank exteriors for cracks.

"The cameras operate at a measuring frequency of 72ms," says Endre Toth, Director of Business Development for Vision Component's U.S. office in Hudson, N.H. "The overall cycle time of the system depends on the number of individual measurements per true running test. In this particular application case with 15 measurements, the system reaches a cycle time of 3.2 seconds. The testing facility provides high measuring accuracy: in the true running and the diameter test, the measurement uncertainty is less than 1µm. It ensures reliable, precise and fast quality control—such extensive checks could not be realized by means of conventional mechanical methods."

Another good application for smart cameras is the inspection of glass pipettes for cracks and dimensional tolerance. Using DALSA's iNspect™ vision software on its VA30 vision system, one camera has a magnified view of the pipette tip and can detect a crack in the glass tip, Dawson explains. A second camera has a larger field of view and is used to measure the pipette's length and critical diameters. The entire system is easily integrated into the customer's existing production line."

INTEGRATION

Generally, integration of a smart camera is easier than a larger automated system. "Remember that a smart camera is a single unit, so it eliminates the need for several components [that can come from different vendors]," says Matrox's Perelli. "There's no need to assemble the components—image acquisition hardware, CPU, camera and optics. Dealing with a single vendor means a single point of contact. Should something go wrong, it will be much easier to pinpoint the location of the fault. Furthermore, if the function of the camera ever needs to change, it can be brought back to the PC and reprogrammed with the new information."

Some of the high-end cameras are application specific, says JAI's Kinney. "In these cases, within the limits of the application—barcode, simple OCR, some semiconductor types and simple inspections—the products may be easier to use. However, if you are on the edge of what they were designed for, have challenging parameters like limited lighting, or other factors, their performance becomes marginal, they have limited options and they may actually become harder to use. If it doesn't work immediately, it may be harder to get working, harder to identify what is wrong or which parameter needs modifiying, which can consume a lot of time. Thus, one of the sources of hidden costs."

THE FUTURE

Our experts say smart cameras are here to stay.

Michael Schwaer, Basler: "Smart cameras have their future in areas where a small, self-contained, powerful solution is required. Because they are easy to integrate, make less demands on the infrastructure of the environment and are able to evaluate sophisticated althorithms, they are becoming more and more relevant in an increasingly cost-conscious business environment."

Steve Kinney, JAI: "It is notable that the AIA has recently broken the smart camera market into three segments: 1) essentially the fully integrated smart sensor segment, 2) the open platform traditional smart cameras, and 3) a vision appliance [small, standalone, dedicated processor with I/O] connected to a camera. It is the addition of the third category that I think is the most interesting and more the wave of the future."

Ben Dawson, DALSA: "Smart cameras will become faster, smaller and easier to integrate. Their software will improve and their applicability will expand beyond manufacturing quality control."

Fabio Perelli, Matrox: "Certainly we can expect CPUs to shrink and emit less heat; this will broaden the types of applications where smart cameras can be used."