Bewildered and adrift

Universities get a "C" in teaching practical machine-vision-system design, and more help is needed from a recalcitrant industry.

by Andy Wilson
EDITOR
[email protected]

Anyone new to the field of machine vision may at first be confused by the variety of hardware and software choices available. Certainly, the industry is a long way from the plug-and-play computer graphics, industrial networking, and wireless markets in which near-commodity products have allowed systems to be rapidly deployed. But it's not just the plethora of nonstandard cameras, optics, frame grabbers, and lighting devices that system engineers are facing. There are also a number of software choices ranging from off-the-shelf imaging packages to proprietary operating systems and development tools.

All this, coupled with the need to integrate host PCs, graphics, networking, and motion control, can be a particularly daunting task for the systems engineer without a Ph.D. in computer science and image processing and mechanical engineering! Last month, we spoke to Benjamin Bachrach of Intelligent Automation Inc. (Rockville, MD, USA), who admitted that in his 80-person company there were 20 people with Ph.D. degrees and another 20 with M.Sc. degrees (see Vision Systems Design, March 2004, p. 21).

So where do you turn to find hope in this bewildering sea of image processing and machine vision? Certainly, here at Vision Systems Design, we like to think that the systems-integration examples we highlight help you to make more productive decisions. But we cannot go it alone. Several of our readers come from mechanical-engineering backgrounds, while others may focus on computer science. Because of this, many companies require a multidisciplinary team when faced with designing a machine-vision-based manufacturing-automation system.

Unfortunately, the educational systems here and in Europe are mainly to blame. Certainly such institutions are good at teaching the basics of computer science, mathematics, and design engineering. Every subject is very neatly divided up: Calculus I, Calculus II, Computer Science I, Computer Science II, Mechanical Engineering I, Mechanical Engineering II, Image Processing I, Image Processing II—you get the picture. But after three years or more of this, the student is left to wonder how it all fits together.

Rather than explain this, many university professors are content to continue in their old ways of turning up at class, giving a lecture, and leaving. And in their spare time, they prefer to write textbooks that outline all these basics. Unfortunately, the old adage that "if you can't do it, you teach it" remains prevalent.

While their basic courses on fundamental mathematics and computer science remain sound, universities could augment these subjects by offering more system-design courses taught by knowledgeable people from industry who have experience in developing practical applications. And instead of writing textbooks on just the basics of a specific subject, more effort could be placed on books that show how such integrated systems can be practically implemented.

Although mainly the fault of the educational system, others—including trade publications, professional associations, trade-show organizers, and manufacturers—need to take some of the blame. Trade publications often republish unedited press releases or publish articles that are little more than marketing hype. Trade-show organizers, driven by the hope for profit, often throw together "sessions" where marketing managers promote products. Few manufacturers are involved at all in continuing education.

There is hope. Last year, the Automated Imaging Association (Ann Arbor, MI, USA; www.machinevisiononline.org) held a very constructive seminar on machine vision and robots that more should have attended. They should be congratulated for their efforts.

But much more needs to be done. And we can't just sit around and wait until everything becomes "plug-and-play." In machine-vision-system design, it will never happen.