Innovations in wavelength and pulse control allow LEDs to be more efficient in industrial settings, and developers say lighting should be considered at the start of system design.
HANK HOGAN, CONTRIBUTING EDITOR
For potent implementation of light-emitting diodes (LEDs) and light sources in machine vision applications, illumination needs must be integrated into the initial system design. And because image processing is one of the key functions of these systems, it is important to remember that the specific technique of illumination has a direct bearing on the usability of an image in quality inspection, process information collection, and other tasks, said Michael Stelzl, managing director of the machine vision solution development supplier MSTVision.
Stelzl said that the success of a machine vision application is driven by the quality of the image that informs the system in the context of a particular process, and illumination plays a fundamental role in providing this quality. He said that shortcomings in an illumination system are likely to produce an image that lacks sufficient contrast to reveal critical details. An inadequate image can make it difficult or impossible to determine whether a part meets specifications in an inspection process, for example.
Time-division multiplexing of different LED illuminations, performed as part of a battery inspection. Courtesy of MSTVision.
Advancements in LEDs are enabling them to be employed to meet machine vision lighting needs. LEDs are available in more wavelengths than ever before, with capacity from the ultraviolet (UV) to the infrared (IR) portion of the spectrum. LEDs are also more efficient, producing more light per watt of power than traditional light sources. According to Fabien Dubois, applications engineer at LED illumination manufacturer ProPhotonix, the latest LEDs are ~50% more efficient than older iterations of these light systems and can also be driven at a higher current.
Wavelength range
The power efficiency available for
an LED system depends on the specific wavelength in which it is operating.
In terms of what is currently commercially available, a blue LED that produces light between 400 and 480 nm is approximately an order of magnitude more efficient than one that emits in the UVC band, between 260 and 280 nm. Today’s LEDs that are applied in the shortwave infrared region, between 1000 and 1700 nm, also do not perform as well as those that emit in the visible range.
LEDs contain gallium nitride, indium gallium arsenide, indium phosphide, or another compound semiconductor. The semiconductor composition dictates the LED output wavelength, although the use of a phosphor in the optical path can shift and broaden the emission spectrum. By devising new semiconductor compound recipes, researchers are developing the basis for LED emission wavelengths that can be applied for machine vision.
Pulse settings
In addition to expanding the available wavelengths, another focus of innovation among developers is the enabling of shorter pulse duration. Having an illumination source turned on only when it is needed reduces waste heat and cooling requirements. Pulsing LEDs and reducing their duty cycle can extend their lifetime, increasing the uptime of a machine vision system.
LEDs enable computational imaging to extract 3D and other data by illuminating an object from different angles (four directions in this example) and using different lighting conditions. Courtesy of CCS America.
Shorter-pulse LEDs can also improve machine vision performance for inspection or analysis. Parts moving on a conveyor belt, for instance, only need to be illuminated for an instant, said Yves Bertic, senior business line director for visible products at Luminus. However, it is often an advantage to have as much light as possible during the brief time that the LED is on in this scenario. A shorter pulse width — when compared with a longer one or continuous operation — enables a more intense burst of light.
“You can drive it at a higher current without damaging it,” Bertic said. Materials and other packaging parameters that are selected during development and manufacturing can be tailored to heighten the ability of LEDs to safely handle high-current conditions, he said.
The gating factor in pulse width has traditionally been the camera, which requires a certain amount of time to capture an image. Advancements in the capabilities of modern cameras have reduced the exposure time, leading to a demand for the integration of faster-switching LEDs.
Dubois said he and his fellow engineers have seen shorter pulse times applied especially in multispectral imaging, in which lights are flashed sequentially.
Packaging innovations improve LED capabilities. Wire bonding of the LED die in the chip-on-board manufacturing step connects it electrically to the package. Courtesy of ProPhotonix.
He said that a multispectral or hyperspectral system might have seven, eight, or more wavelengths. Each wavelength provides illumination
that covers a part of the spectrum,
and combining images generated with different wavelengths creates one picture that spans the entire spectral range.
Latest in packaging
In addition to wavelength and pulse duration, important developments have occurred in the size and performance of LED packaging. Daniel Huber, director of LED lighting supplier TPL Vision, said that surface-mount LED technology, which provides uniform illumination, is getting smaller and smaller and enabling a higher light output per area.
At the same time, vendors are rolling out chip-scale packages. Fifteen years ago, a high-power LED die might have measured a tenth or less of the total area of its package. So, 90% or more of the surface would not produce light.
Today, with chip-scale packages, the LED die is almost the same size as the body of the LED package, allowing a larger portion of the surface to generate light. This development, in turn, increases the amount of light available for machine vision applications. The reduction in the size of the chip and package also makes it easier to fit a lighting solution into a confined space — a requirement in many machine vision applications deployed in tight confines on a factory floor.
Multispectral sorting using a monochrome shortwave infrared line-scan camera. Such multi- and hyperspectral imaging is a trend, creating new lighting demands. Courtesy of MSTVision.
These packaging developments combined with broader spectral output yield benefits, Huber said. “These advancements in lighting technology provide more accurate and natural color reproduction, making it possible for machine vision systems to perform more precise visual inspections,”
he said.
For the best results in a specific machine vision application, the lighting may need to be bright, at a particular wavelength, from a certain direction, in a specific and complex sequence, or some combination of these characteristics. An application in which all these factors come into play, for example, is the reading of product-identifying text on a tire. The challenge, from the standpoint of imaging, is that the text may be raised from the surface but is otherwise the same color as the surrounding material. Hence, the contrast is low, making it difficult for a traditional machine vision system to capture the relevant information.
Power of computational imaging
The potential solution to fully resolving views with inherently low contrast lies in computational imaging, said Jason Allen, engineering manager and precision LED lighting specialist at CCS America. The computational imaging process begins by taking pictures of the object under several lighting conditions from different directions. Then, it is possible to extract 3D data and other information from the images, collectively, with the aid of software.
“Those images are all processed into a single image that will show you what you want to look at. You’re using the light to do shape from shading for photometric stereo,” Allen said.
Light that’s just right
In discussing these and other challenging applications, Allen said that projecting more light on an area
may not be the answer. Flooding a
field of view with light may wash out fine details, and what ultimately constitutes too much light depends, to a degree, on how big the field of view is. A small space may be overwhelmed by a level of illumination that would not be a problem in a larger space. Hence, choosing the right lighting setup requires careful consideration of several factors.
The best time to evaluate options for a particular lighting solution is during the design stage of a machine vision project, according to Torben Deike, global product category manager for Basler. The company develops industrial cameras and partners with suppliers of other vision components to achieve optimal interoperability. The goal of such a collaboration is a system that has better overall performance than multiple independent systems and feels like a single product to the user, Deike said.
He said that Basler cameras feature intelligent light control. This capability allows the camera to periodically increase the brightness of the light source over the machine vision system’s life cycle to compensate for LED aging and ambient light changes.
A photometric stereo line scanner, applied to the inspection of surfaces in electric vehicle battery manufacturing. Courtesy of MSTVision.
“If the lighting control is a standard feature of the camera and Generic Interface for Cameras (GenICam) compliant, such effects can be more easily compensated because everything is controlled via a single software, simplifying synchronization and parameterization,” Deike said.
Given the variables inherent in any machine vision solution, Allen advises that anyone trying to decipher which LED and illumination will produce a suitable and consistent image should test out the light solution prior to purchasing.
For this purpose, CCS America will set up a demonstration in its own lab of a proposed machine vision solution, he said. The company will also loan out lighting setups for trial runs in a customer’s location.
Deike’s advice when it comes to lighting is to use industrial-grade components, as opposed to commercial-grade components. Doing so ensures more consistent lighting properties through special binning procedures. Considering the integration of lighting as early as possible in the design of a project makes it less likely that a custom lighting solution will be needed, he said. This early planning will enable the use of more readily available standard components to be integrated into the machine vision system.
Braille dots on pharmaceutical packaging, an example of an application illustrating lighting challenges. Courtesy of MSTVision.
Of course, sometimes a customized lighting solution is necessary due to its specific task. For example, a specialized need was identified when MSTVision set out to address high-speed scanning applications. For this purpose, the company often uses a multiplexed time-division approach, firing a sequence of high-intensity lights so that the composite of the pictures yields a clear image, Stelzl said. MSTVision developed its own photometric stereo line light when it was unable to find a commercial product that could meet performance requirements.
One application for photometric stereo line light is Braille printing in pharmaceutical applications, which requires 100% inspection to ensure that it is
accurate and defect-free. In developing its solution, MSTVision used high-intensity LED flashes. A continuous lighting solution would have consumed much more power, significantly decreased system lifetime, and necessitated water cooling — an unsuitable trio of requirements in such settings, according to Stelzl.
Looking to the future, what Stelzl
desires in lighting solutions can be boiled down to a key metric, one that applies to many applications and
therefore could drive a trend within
LED development. “I want to have as much power as possible per area,” Stelzl said. “For me it’s power per emitting surface that needs to be more and more and more.”