Sleeps with the fishes

Dreadful things may happen to vision systems designers who fail to find components that stand the test of time

Vision systems are used in a number of applications, and those that emerge from the labs of universities are perhaps some of the most exciting. Last month, for example, I discovered that many researchers are studying how to embed machine-vision and image-processing algorithms into autonomous robots. One of the most interesting concepts was to embed the technology into autonomous robotic fish that would swim in the ocean and allow the behavior of fish to be remotely studied.

After building the robotic fish, the researchers in question successfully deployed the device to remotely transmit images of fish behavior to a surface vessel. The robotic fish was indeed very lifelike and useful in producing research data; unfortunately, the developers had overlooked one important fact.

It was so lifelike (at least to other fish) that the $20,000 robot was eaten by a shark. Needless to say, the researchers were not amused. What they had not taken into account in their desire to discover more about fish behavior was how quickly the semi-autonomous system would be rendered inoperative.

Although engineers of image-processing and machine-vision systems do not face this type of extreme scenario, many factors must be considered to ensure that their systems do not become obsolete early in existence.

Today, many designers choose to use off-the-shelf components when developing their machine-vision systems. Granted, the low component cost proves beneficial to developers, but the large number of small and medium-sized OEM suppliers of such products can also prove detrimental, especially in a recession.

For instance, imagine you have specified a system that inspects rotating parts as they move along a production line using a number of different lighting peripherals, cameras, frame grabbers, I/O boards, controllers, and software. After building and shipping this system to the customer, you find to your horror that one of the component suppliers has ceased production of the part due to lack of market demand—or worse, closed its doors.

Know thyself—and thy supplier

Though many off-the-shelf OEM components such as computers, monitors, or I/O peripherals used in industrial automation systems are interchangeable, those from manufacturers of machine-vision components are generally not.

LED lighting manufacturers may supply dome lights at similar wavelengths, but these may vary in intensity depending on which current controller is used. A high-brightness LED back-panel from a less-than-well-known vendor may provide the contrast required to silhouette an image, but there may not be any products available that could even come close to offering equivalent performance. Worse, if these lights are overdriven, the effects of luminosity degradation will degrade the performance of the system over the long term.

This represents just one example of a worst-case scenario. Any similar event can have dire consequences in terms of both engineering time and money spent on a redesign. To ameliorate the chances of such a catastrophic event, developers must take a very conservative approach in the design of their machine-vision systems.

In system design, specifying the features of individual OEM products is very important. As important, however, is assessing the reputation and quality of the supplier; determining whether any equivalent or near-equivalent components are available; and ensuring that enough equivalent performance margin exists between these products. This way, should your OEM supplier fail to deliver, you will not be faced with an expensive redesign. Otherwise, your next system may end up like yesterday’s newspaper, wrapped with fish and chips.