Industry Insights
Lighten Up, Francis: Filter-Light Combinations Ease Machine Vision Data Requirements
POSTED 09/05/2013 | By: Winn Hardin, Contributing Editor
The right filter-light combination can reduce the processing and data transmission burdens on your machine vision system.
If you need to find a green tab on a label, sort pills by color, or separate a black barcode from nearby red type, you might reach for a color camera as the front end of your machine vision solution. But if your inspection key is a single color, not a shade of a certain color, you might be adding extra work and cost to your machine vision solution. Using the right color filter with the right illumination source might make a successful solution easier and cheaper.
“Packaging applications use a lot of filters because the usually involve color of some kind, but the customer may not want the expense of a color camera solution,” says Matt Pinter, Design Engineer at Smart Vision Lights (Muskegon, Michigan). “Color cameras have their advantages, but if you can use a green light with green bandpass filter to find a green tab – basically make everything green appear bright in the image and everything else appears dark – then you eliminate the cost of the color camera, the additional data burden of sending color image data, and reduce your communication requirements and your processing requirements. But the color you’re seeking must have a definable range. Experienced vision professionals understand the interaction of light with objects and the effect on imaging systems. Seasoned vision pros love filters.”
Filter Options
“By far, the most popular types of filters we sell are bandpass – the color filters that allow green, blue, or red to pass through them and block the other wavelengths of light; the short- and long-pass edge filters that pass only light below or above a certain wavelength, respectively; and polarizing and neutral density filters,” explains Nicholas James, Imaging Product Line Manager at Edmund Optics (Barrington, New Jersey). Polarizing filters block light waves that are not aligned with the given angle around the center of the electromagnetic wave. This type of filter is particularly useful for filtering out reflections from metal and other polished surfaces. Neutral density filters block a certain percentage of light across a broad spectrum. These filters are regularly used around very bright light sources, such as welding machines and metal forging.
Out of these filters, bandpass filters are the most commonly used in machine vision applications. “The main reason vision designer’s use bandpass filters is to enhance contrast or highlight a feature,” Pinter explains. “In fruit applications, for instance, using an infrared light with an infrared bandpass filter allows a black-and-white camera with sensitivity in the near-infrared to see bruises that wouldn’t be visible otherwise.”
“By correctly matching the filter, light, and reflective properties of the object, you can potentially get more useful information into the system without having to spend as much time on programming,” James adds. “The more you can do without software and post processing, the easier it is on the vision system, which either reduces costs or allows you to increase throughput. This is particularly important as the industry moves towards higher-resolution cameras that also add to system’s data handling requirements.
“The general rule is ‘Like colors lighten like colors,’” he continues. “A red light with red bandpass filter will highlight a red object compared to other features that are not red, in other words.”
Both bandpass filters and edge, also called short- and long-pass filters, can also be used to protect equipment and workers. “We often focus on improving contrast, but in various semiconductor processes like lithography or dicing, for example, you’re often interested in removing certain wavelengths to protect photosensitive materials,” explains Jason Baechler, Business Manager, Industrial at SCHOTT North America, Inc. (San Jose, California). “If the laser used to dice a wafer reflects back to the camera, you can saturate your image or damage the camera, for example. So you might want an edge or a notch or band-stop filter there to block a specific wavelength of light. Vision designers also regularly use infrared light with infrared bandpass filters that block out ambient light and reduce the strain of visible strobe lights on nearby workers. Flashing lights for hours on end can cause eye strain on nearby workers or even result in seizures for workers with certain medical conditions.”
Light Lenses
Just as filters can be used to shape the available light spectrum at either the light source or the receiver (camera) to improve system performance, optics at the light source can also improve system performance for relatively low cost compared to other machine vision components.
“Our back lights are the only light we produce that doesn’t always include some type of lens,” says Smart Vision Light’s Pinter. “In many cases, it’s better to carefully choose the light-lens combination to provide even illumination over a given area than to use a diffuser because diffusers absorb and scatter a lot of the light’s output.”
LED lights typically use cost-effective, total-internal-reflection (TIR) molded plastic lenses that add around $1 to the total product cost. Common lenses spread the light 10 or 30 degrees, but there are options for spreading the light up to 70 degrees or more. While diffusers are often used to improve the light uniformity across a scene, Pinter adds that careful selection of lenses and the position of individual LEDs within the light can achieve the same goal.
When approaching a new machine vision system design, experts say it is important to work with a distributor or supplier that has lights, filters, and lenses at their disposal to try out different combinations and find the best set for a given application. Today, trial and error mixed with experience still represents the best path to success. Unfortunately, optical models do not exist for every optical component and object under test, which means designers cannot rely on complex optical design and ray tracing software to automatically solve every machine vision design challenge.