Edited by Miles Budimir
All too often, machine-vision systems are built that either fail to meet performance expectations or are chock full of overspecified components. Underspecified systems tend to fail and must be redesigned, whereas overspecified systems are often more expensive than needed.
It's easy to avoid these pitfalls. Pay attention to the specs. They should be based solely on the application. Vision systems extract information from an image. How the system will be used determines the required image quality. Will the system have to capture images quickly? Does it need to check the orientation, color, or size of the workpieces, or merely detect presence? How large is the smallest detectable detail? And how much contrast is necessary?
Image quality depends on the quality of the components. Thus it cannot be specified by a single spec but by several factors. The fundamental parameters for image quality are:
1) Field of view (FOV): The viewable area under inspection. In other words, the portion of the object that fills the camera's sensor.
2) Working distance: Distance from the front of the lens to the object under inspection.
3) Depth of field (DOF): The maximum object depth that can be maintained entirely in focus. DOF is also the amount the object can move without refocusing and still be in acceptable focus.
4) Sensor size: The size of a camera sensor's active area, typically measured in only the horizontal dimension. This parameter helps determine the proper lens magnification needed for a specific FOV. The primary magnification (PMAG) of a lens is the ratio of sensor size to FOV. Although sensor size and FOV are fundamental parameters, it is important to realize that PMAG is not.
5) Resolution: Minimum detectable feature size of the object under inspection.
Resolution measures an imaging system's ability to reproduce an object's detail. For example, imagine a pair of black squares on a white background. If the squares are projected onto neighboring pixels, they appear as one black rectangle. A certain amount of space is needed between them if the system is to see them as two objects. Determining minimum distance yields the system's limiting resolution. The relationship between alternating black and white squares is often described as a line pair. Typically, resolution is defined by the frequency measured in line pairs per millimeter (lp/mm).
Closely related to resolution is contrast. It describes how well a system can distinguish black from white. In practice, black and white lines blur into grays. This blurring reduces contrast. For accurate performance, a vision system's ability to reproduce object contrast is as important as reproducing object detail (i.e., resolution). The lens, sensor, and illumination play key roles in determining the resulting image contrast.
The resolution and contrast of an image can each be defined individually, but they are also closely related. Consider two dots close to each other and seen through a lens. Because of the nature of light, even a perfectly designed and manufactured lens cannot fully reproduce an object's detail and contrast. When the dots are far apart (in other words, at a low frequency), the dots are distinct. As they approach each other, the dots blur and overlap until individual dots can no longer be distinguished. Resolution depends on the imaging system's detecting a space between the dots. Therefore, optical engineers usually specify a contrast level at a specific resolution.
The relationship between resolution and contrast can be expressed in a metric called the modulation transfer function (MTF). An MTF curve can be generated by plotting contrast at a range of frequencies.
The high-resolution portion of the curve is not always the most important part of an MTF. For many applications, high contrast at low frequencies is more crucial than the absolute resolution. For such applications, a higher-resolution lens designed to work with film rather than with CCDs, for example, will not improve the overall system, although it will increase cost. More even illumination may be all that is needed.
One of the most common mistakes in specifying vision-system components is to use magnification as the primary parameter. As pointed out earlier, primary magnification depends on FOV and the sensor size. Too often, it is easy to get caught up with magnification and end up going astray.
Often designers develop vision systems after working with microscopes. If they were using 20X magnification, they assume this will work in a machine-vision system. But the fact is the two magnifications are not the same. Microscope magnification takes into account the eyepiece and the user's eye. The PMAG in machine-vision systems don't. Also, the sensor size figures into PMAG, a parameter that simple microscopes don't have.
Another pitfall is illumination. Using the best imaging components will not guarantee good image quality. Even if contrast and resolution issues have been solved, proper illumination often gets overlooked.
When to go custom
Sometimes, an off-the-shelf lens just won't cut it. In those cases, consider a custom lens. Even though they are more costly and require more design time, as well as long manufacturing lead times, there are cases when custom lenses make sense.
Going to a custom design can save money in the long run, especially if the lens will be made in high volumes. And because the lens is designed for a specific application, adjustments normally built into an off-the-shelf lens can be eliminated. For instance, if there are unusually tight requirements for working distance and packaging, a custom lens may be the only option. Also, if a specific FOV is needed, a fixed-lens design may be less expensive than a zoom lens that includes the right field of view but with a lot of unused adjustability. Or if the illumination is constant, the iris can be fixed at one setting. If several lens elements are called for, the costs of a custom design and set-up can be amortized, reducing repeating costs like the number of lenses and mechanical adjustments.
Off-the-shelf elements can be incorporated into custom lenses. Look first to off-the-shelf components to make design, prototyping, and production faster and less expensive. A simple and cost-effective way to work off-the-shelf lenses into designs is to choose the lens before the mechanical design of the system is finished. Waiting until the housings are set makes integration much harder. When designs are more demanding, completely custom designs may be necessary.