A comprehensive imaging reference

Steve Scheiber, Contributing Technical Editor -- Test & Measurement World, 5/1/2005

To help you get up to speed on a variety of machine-vision topics, Edmund Industrial Optics has assembled a collection of application information into its 2005 edition of The Best of EO Application Notes. You can download a copy of the entire book from the company's Web site. Here are some highlights:

Anyone building or buying a machine-vision system has to understand the fundamental image-quality specifications (see figure) and how they interrelate to define the system's overall performance. The goal is to get sufficient image quality to meet your inspection needs without overbuying.

This note explores many of the obvious and not-so-obvious factors that determine the suitability of a particular piece of hardware to a specific application. For example, it defines depth of field—an often overlooked specification that in printed-circuit board inspection can ensure that the tops of tall devices, device leads, and traces on the board surface all remain in focus in a single image. Sensor size is important to determine the appropriate lens magnification required to produce a desired field-of-view.

The fundamental parameters of an imaging system, including the resolution, field of view, and depth of field. Courtesy of Edmund Industrial Optics.

The discussion also explores the impact of contrast, perspective error, and distortion on overall image quality. Although most data sheets specify resolution and contrast separately, they are closely related. Contrast—expressed as gray-scale or signal-to-noise ratio—describes how effectively an image reproduces differences between boundary areas relative to one another. Although its meaning is usually well-understood, its importance to image accuracy is often underestimated.

The note's author contends that resolution can be meaningless unless it is defined relative to a specific contrast and vice versa. Consider an image consisting of two dots. If the dots are far apart, the distinction between them is unambiguous. As they get closer together, the ability to resolve them rather than perceive them as a single object depends both on the resolution of the camera and the level of contrast between the dots and their background. The author further explains how you can use the modulation transfer function, which describes the relationship between contrast and resolution, to determine a vision system's expected performance.

Another discussion of vision fundamentals, this piece serves as a kind of "cookbook"—a useful checklist of considerations for specifying the lens of a vision system. Overall, the note aims to provide you with information about lens characteristics so you can assemble a vision system that provides the best possible performance at the lowest possible cost. For example, the author recommends specifying field-of-view rather than magnification to ensure that the vision system can inspect the entire region of interest. The author also suggests considering factors such as illumination integration, charge-coupled device (CCD) format, operator error, and software development to help reduce setup costs and system downtime while optimizing reliability and repeatability.

The paper also explores factors that affect lens distortion and suggest ways to minimize their impact. Distortion is an optical error that causes differences in object magnification at different points in the image. There is no loss of information, but it is "misplaced." Software can compensate for lens distortion before analyzing or producing the final image.

The author also suggests that, despite manufacturers' reluctance to share it, you should try to get as much information as possible about the lens's design criteria. In particular, inquire about the optimal distance from the object plane to the lens. If you need a lens for a short working distance, you don't want one that was designed to focus at infinity.

This note, one of several in the book that discuss specific applications, relates how one manufacturer required a system to inspect the prototype of a hardware computer key connector to verify the placement of its pins. The author describes the customer's system requirements and shows how to select the appropriate components to meet them.

Because conventional lens designs suffer from perspective errors when imaging objects with significant height or depth, the note recommends telecentric lenses to correct the problem.

The author of this piece contends that planners sometimes begin working on the design of the optical system only after completing mechanical and other system designs. Therefore, the vision system must fit into whatever space is left.

This note outlines a series of steps you can take in order to assemble a workable imaging system in a too-small box:

1. Define your mechanical constraints;
2. Define your fundamental parameters;
3. Lay out the imaging system as though it were in a straight line;
4. Place the illumination and determine the minimum f-stop;
5. Compare the optical design with mechanical constraints; and
6. Bend the system.

To determine the mechanical constraints, for example, you must first determine how much space is available for the imaging system. Determine the size of the box the system must fit in and measure the length of the space allotted for optics. Remember that the space will have to accommodate the length of the camera lens and the length of cables and, in most systems, the illumination components.

The author recommends that you first lay out the system in a straight line to ensure it works before fitting it into the allotted space. If the system doesn't fit, then you may need to introduce "bends" into the optical path with prisms, mirrors, or beam splitters. The author briefly discusses the pros and cons of each option and shows how an effective optical system can be designed for a space that initially seemed too small to hold it.