This last decade has seen a significant increase in the popularity of food-processing imaging systems to perform tasks such as quality assurance, food-safety inspection or process control. Nowadays, almost all major food-processing enterprises have some kind of vision system in operation in their facilities.

One thing almost universally required of these systems is a uniform and stable lighting configuration. Researchers at the Georgia Tech Research Institute (GTRI) continuously evaluate advancements that could improve the performance and utility of delivered systems. In this case, a study was conducted to evaluate the performance of the latest Light Emitting Diodes (LEDs) for industrial imaging illumination.

The early vision systems developed by GTRI used high-frequency fluorescent lights, as they were more efficient and provided relatively uniform lighting distributions. Although they are more efficient than incandescents, fluorescent bulbs generate enough heat when enclosed and operated in industrial settings to require cooling in order to achieve the desired light intensity stability. This added need for cooling drives up costs and introduces an additional critical failure point for the imaging system.

In 2004, researchers began experimenting with LED lighting, and in 2005, they fielded the first prototype system using LEDs. Although the LEDs are typically more expensive than fluorescents, they provide benefits such as longer operational lifespans and strobing functionality. Food-processing operations typically have relatively low throughputs; for example, broiler shackle lines run at approximately three birds per second.

For an imaging system designed to capture images of each bird, the LED illumination can be strobed to only be “on” during the image acquisition, much like the flash on a camera. With a typical integration time of 3 milliseconds per image acquisition of each bird, the LEDs would be on for only 9 milliseconds of each second, or roughly one-hundredth of a second!

Strobing the LEDs in this way also significantly reduces the amount of heat generated by the lighting system, reducing the need for additional cooling.

The discontinuation of the original LEDs necessitated the identification of a suitable replacement for the illumination in both future imaging systems as well as in existing fielded systems. The research team reviewed several LED specifications and selected a tentative replacement for validation. Although the team expected to find better efficiencies and tighter color specifications, they were impressed by the significant improvement in performance just in the last few years.

Multiple tests were designed to compare the obsolete LEDs (Luxeon Batwing LXHL-MWJE) with the new replacement (Cree MPLEZW-A1-R100). The evaluation procedure included: (1) establishing the worst-case efficiency for both LEDs, (2) measuring the peak reflected illumination intensity using a high-precision amplified silicon detector (Thorlabs PDA 250), and (3) comparing the spatial illumination profile and color conformity from collected image data.

The efficiency of the LEDs was calculated from the given specifications as the light output in lumens per unit of electrical power. A lumen is a measure of the power of light as seen by the human eye.

In this case, an obsolete Luxeon LED outputs 450 lumens with a forward voltage of 24VDC at 1050 milliamps. A candidate replacement Cree LED generates 800 lumens with a forward voltage of 26VDC at 450 milliamps. Calculating power, the Luxeon LED consumes 25.2 watts compared to the Cree LED’s 11.7 watts under the same environmental operating conditions. Using the above numbers, the calculated efficiency is 17.9 lumens per watt for the old Luxeon LED compared to 68.4 lumens per watt for the new Cree LED, indicating an increase in efficiency of 283 percent.

Performing a similar calculation to compare light output vs. purchase cost also shows a cost improvement of 390 percent when moving to the new Cree LED. Not only is the new LED far more efficient, it is also effectively less expensive. A chart illustrating this comparison is shown in Figure 1.

Reflectance tests were carried out to empirically compare the light output of the two LEDs. The test setup consisted of the LED being tested, a PDA 250 silicon photodiode, and a white reflectance standard placed at a fixed distance and angle.

Reflectance data was collected for each LED at current levels ranging from 750 to 1.5 milliamps. Figure 2 shows the results, clearly demonstrating that the reflected light from the Cree LED is significantly higher than that of the Luxeon LED across the range of operating currents. In fact, these tests showed that the reflected light intensity from the Cree LED is more than 400 percent brighter than the reflected light intensity from the Luxeon LED at the same operating current.

Other important attributes of illumination sources are the color conformity and the spatial illumination profile. Imaging systems are often used to accurately classify product color for operations such as oven control and defect detection.

Color conformity was measured using a color camera to capture an image of a MacBeth color tile board illuminated by each LED operating at 750 milliamps. Figure 3 shows these two images and the RGB color profiles across the top row of tiles in the image.

The difference in overall brightness between the two images can clearly be seen. The similar color profiles indicate that algorithms designed for use with the Luxeon LED should not be significantly affected by a transition to the Cree LED as an illumination source.

With new applications, such as automotive lighting, spurring consumer and industrial demand, LED manufacturers have responded by making LEDs much more powerful, more efficient, less expensive, and more flexible.

This has made LED lighting the illumination source of choice for most imaging and machine vision applications.