- This article is filed under:
- » View All
What Every Machine Vision User Should Know About Lighting
by John J. Merva, Machine Vision Consultant - AIA Posted 02/12/2002
As experienced vision users will tell you, lighting is always a critical component of a machine vision system. Nonetheless, many vision project engineers choose the light as the last step of component selection. Not only does this jeopardize the success of the project from the standpoint of achieving the desired results but also often results in the purchase of components that ultimately will not be used for the project. Following are some tips that will help you succeed when selecting lighting for your application.
1. Include the light from the start - Like all systems, vision system components need to be designed and selected based on the strengths and weak points of other parts of the system. Selecting any without the others in mind can lead to serious design shortfalls.
2. Try not to start with your toughest project - This is a good piece of general advice for all machine vision projects but stands particularly well regarding lighting tasks. As a mathematician friend of mine is fond of saying, 'Everything is easy once you understand it'. Illumination optics requires a different way of viewing the problem than you may have developed for projection or imaging optics. If possible, avoid difficult lighting and imaging situations until you master the basics. If you must solve a difficult application, use outside resources to help.
3. Lighting is the least expensive processing power you can buy - Reliably detecting flaws or measuring features relies on creating contrast between features of the part. Sometimes referred to as segmentation, this can be accomplished through sophisticated algorithms running on the vision processor. If you utilize good lighting analysis principles, much of the image segmentation can be accomplished through correct illumination specification. A few hundred extra dollars invested in the right light can save you thousands of dollars of machine vision system.
4. Be from Missouri - People from Missouri are known to have the show me before I will believe it attitude. You should make your lighting vendor show you before you believe they have a solution for you. All machine vision vendors are now prepared to offer images of your part as it will appear digitally in the system. In fact, most vendors offer this as a free service to demonstrate their capability when solving a customer's application. If your vendors can't provide images of your sample parts that effectively provide contrast where needed, they don't have the solution you need. Remember, illumination optics is a science not magic.
5. Do your installation design after you choose the light - Many times the camera mounting fixtures have been designed and built before the lighting solution is specified. This often creates a no-win situation. One of the primary components of a lighting solution is the geometry of the light field applied. Pre-designating mechanical space before the proper solution is identified often precludes the use of the correct lighting. Original Equipment Manufacturers of machinery should really pay attention here.
6. Move from wave and look to understanding the science - As mentioned earlier, illumination optics requires the practitioner to form a different paradigm for viewing the problem. Most of us have developed lighting skills more like artists than scientists. Artists often work from feel; scientists work by following a process to impart rigor and understand new situations based on knowledge already documented. As a new vision applications engineer, I was very effective using the 'Wave and Look' method for lighting selection. To utilize it you set-up your part in the correct camera field of view and try different lights until you see the image you want on the monitor. When you get the right image, you design a mount to hold the lights where you positioned them. While often effective, this technique sometimes yields less than optimum results. The Automated Imaging Association and lighting and vision companies now offer training in the science of machine vision lighting. Understanding how different surface finishes interact with geometry, marking methods, materials and colors will allow the specification of a lighting solution that will bring the most reliable and repeatable results for the least money.
7. Understand the difference between bright field and dark field - Bright field and dark field are two terms that you will often hear your lighting vendor use. To your amazement they may make small changes to the light position and declare they have gone from bright field to dark field or vice versa. How do you understand this apparent sleight of hand? First of all, while bright field and dark field are used to describe lighting in many applications, they only apply where specular (shiny) surfaces are involved. The second thing to remember is that bright field or dark field is the result of the positioning of the viewing point (i.e., camera and lens combination) and the position of the light. Changing any of these components can change whether your application is in bright or dark field.
Here's the easy way to determine which type of light field you are creating. Place a mirrored surface where the part to be inspected would be. Set up the camera, lens and light exactly as you would for the inspection. If you see the light source reflected in the image of the mirror, you are in bright field. If you don't see the light source, you are in dark field.
What if you see light in some parts of the image and not in others? This means that you are in incomplete bright field. When imaging specular surfaces this situation is often the root cause of poor quality images. Bright field areas in otherwise dark field images are usually considered as glare, an unwanted imaging artifact. This is often the sign that you need a specialty diffuse lighting product.
Figure 1 - Understanding Bright Field and Dark Field Lighting Geometry
8. Learn the strengths of each light source type - With a multitude of different lighting sources on the market, where should you start in determining which to use? The answer is of course the one that is best for your application. Making this determination means that you will need to consider several factors from your project. First of all, almost all lighting geometries can be obtained from all lighting source types. The most commonly used sources in machine vision are fluorescent, fiber optic, light emitting diode (LED) and incandescent.
With cost always being a consideration, many users immediately gravitate towards inexpensive fluorescent ring or LED spotlights. Don't ever forget that your foremost goal should be obtaining a satisfactory image. Remember, the pain of poor performance will remain long after the sweetness of low cost. It won't take too many extra engineering hours to offset any savings made by buying only on cost.
Color and color temperature are often primary drivers in light source selection. Choosing the correct color, or color temperature for color vision applications, is often key to obtaining workable images. Often this aspect alone will lead the user to the proper light source. While fluorescent, halogen and LEDs all can provide white light; you may find a substantial difference in the resulting image of your part when changing from one to another. To understand this topic further, read more on color rendering and color temperature in your favorite lighting texts. Also it is important to remember that LEDs now come in a variety of colors and color filters are available for most fiber sources so you do have many options.
Maintenance and stability of output over time may very well be the next most important considerations. While LEDs are known for long life (up to 10 years of continuous use) and fairly stabile output, they do cost more. However, the cost of shutting down a production machine to replace a light at an unscheduled time may more than offset any savings on the original light purchase. Keep in mind that even LEDs exhibit fluctuation in output due to variations in voltage, current and temperature. If very fine light level control is required, controllers that have feedback loops to control these effects are required.
Fiber optic solutions are often used to allow unique geometries and shapes of lights, which are not available from any other source. This feature allows fiber lights to be used in many applications where other sources can't. Fiber optic power supplies typically use halogen bulbs and can deliver very high intensity while keeping heat remote.
9. Know the difference between diffuse and specular - Understanding this element of surface finish seems straightforward, but in practice, it may not be. Why? It is easy to list examples of diffuse and specular objects. White paper (non glossy) is an example of an almost perfectly diffuse surface. On the other hand, a mirror is an example of a perfectly specular surface. Diffuse surfaces scatter light striking the surface in many directions. Specular surfaces reflect light at the angle of incidence, just like a pool ball striking the bumper on the table. In the real world of part surfaces, many surfaces are not perfectly specular or diffuse, but lie somewhere in the continuum in between. Surfaces such as semi glossy paper or a mirror with a sand blasted surface exhibit light reflection properties of both surfaces. This is often the situation in which you will want to consult someone experienced in machine vision lighting specification.
Lighting is often described as an art. A friend of mine who went to art school often replied that lighting was not taught in art school. Lighting is a science and following an engineering approach to solving lighting problems will yield the strongest solutions. Machine vision users now have half a dozen competent lighting vendors from which to provide lighting solutions. It has never been easier to procure the right lighting solution for your application. Follow the tips above and you'll be on your way to successful machine vision lighting selection.
The author invites questions about this article. Contact him at:
John J. Merva
JJMerva and Associates
68 Oak Ridge Rd.
Weare, NH 03281
Phone: (603) 529-5117
There are currently no comments for this article.
Leave a Comment:
All fields are required, but only your name and comment will be visible (email addresses are kept confidential). Comments are moderated and will not appear immediately. Please no link dropping, no keywords or domains as names; do not spam, and please do not advertise.