Machine vision is a key control element for countless robotics and automated systems, as well as the basis for a wide variety of systems that measure and extract information from objects. But without proper illumination, vision systems are essentially blind.

Lighting requirements for machine vision are relatively straightforward. Whatever's used to illuminate the objects in the field of view must provide adequate contrast for features of interest. It must also be:

  • Bright enough to keep signal levels high and minimize exposure times

  • Uniform so that variations in the object are not masked by lighting variations

  • Stable so that vision-based measurements are consistent

Improper illumination can cause even the best machine vision system to produce dreadful results. Even with the highest resolution camera, most advanced motion control system, and a highly sophisticated analysis program, an unexpected reflection or shadow can render the system unable to locate the objects under test, much less measure them accurately.

Shape up with geometry

One of the first things to focus on when considering illumination is the geometric relationship between light source, object under test, and collection optics. The best geometry for a given application depends on what information is needed from the object as well as its surface characteristics, such as shape, gloss, texture, and color. The International Commission on Illumination (CIE) provides recommended lighting geometries for a variety of surfaces through its website at cie.co.at/cie.

Because of the diversity in application needs, commercially available illumination options span a wide range. Traditional ring lights, dome lights, line lights, and backlights are available off-the-shelf, making illumination setup quick and easy. Besides fiber and LED sources, laser diodes with line-generating optics are common for structured light applications such as surface profiling and distance measurements.

Off-the-shelf solutions also accommodate a wide range of geometries. No illumination geometry is really new, however, so it's likely that a suitable product already exists to satisfy a given situation. Using an off-the-shelf source to solve a difficult lighting situation is often faster and less expensive than creating a custom design.

If an application requires custom illumination, however, lighting vendors can be a reliable source. Vision integrators who have had plenty of exposure to different geometries may have already manufactured a light source similar to the one needed and can create a custom modification. In addition, molds that fit the shape of any object can be made through rapid prototyping to adapt a new application to an existing system.

Color control primer

Once the illumination geometry is established, the next challenge is to focus on the lighting's color and intensity. One of the newest options available is dynamic lighting, which is programmable to provide a wide range of colors, modulation rates, and intensities.

Many sources can be computer controlled to offer dynamic intensity levels and modulated output, but one of the most popular is the light emitting diode type. LEDs are now available that offer high intensity light output in a variety of colors. And because they are solid-state, they provide extended lifetimes.

While individual LEDs typically operate at a single color, some LED-based illumination sources are available with adjustable red, green, and blue output. This control allows the unit to create virtually any color and illumination level required, eliminating the need for multiple lights or additional filters. Such sources can produce not only red, green, blue, and white, but also just about any color in between, including orange, violet, pink, cyan, and yellow. This extensive color range is helpful in obtaining maximum contrast in an image, since most objects are not strictly red, green, or blue.

Troubleshooting tips and tools

Evaluating onsite effects is also important when it comes to illumination. These effects include such things as harsh reflections and changing ambient illumination, which are difficult to predict during system design. Addressing these effects requires a variety of helpful accessories and software algorithms.

Harsh reflections, for instance, may prove to be treatable using linear polarizing filters. By placing a linear polarizer over the source and a rotating linear polarizer over the imaging lens, many types of reflections can be filtered.

If the illumination is at an oblique angle, using a linear polarizer on the imaging lens alone may also work. Unfortunately, using polarizers also reduces the overall amount of light the camera receives, which may cause dark images and problems for machine vision software. Polarization may not be suitable for certain metal objects.

Harsh reflections can also be eliminated by adding diffusion to the illumination. Many different types of glass diffusers are available including opal, ground, and holographic. Holographic diffusers may also aid in shaping the intensity distribution of the illumination.

Several other accessories are also beneficial. For example, using filters over the imaging lens can help eliminate unwanted light or infrared radiation from the ambient illumination of the system's operating environment. Reflectors can redirect unwanted illumination or provide more illumination to certain areas of the object. Long cables that can be stretched and twisted as machine arms extend and contract can help ensure the stability of lighting intensity.

In addition to physical and optical solutions, software offers several ways to correct for lighting issues. Many cameras, for instance, have built-in functions such as “flat field correction” and selectable “knee points.” Flat field correction removes any consistent gradients across an image, such as those that arise from non-uniform illumination. Selectable knee points allow cameras to adjust exposure time throughout the image based on the amount of illumination at each point. Such software solutions are particularly useful when lighting issues can be modeled mathematically.

To read more about machine vision systems, how they work, and how to select critical components, visit motionsystemdesign.com's Knowledge FAQtory and look for links that will connect you to related articles and information.