Picture Picture

The Evolution of Machine Vision

State-of-the-art machine vision systems incorporate fast, high-resolution CCD cameras coupled with intelligent frame grabbers that allow multiple-image processing.


by Michael Neeves

When you hear the words “machine vision,” what do you think of?

If you’re involved in production, most likely you think of an extra set of “eyes,” a flexible go/no-go gage, a quick, unceremonious device that will help maintain production control.

If you’re a metrologist, you may think of a vision inspection system in terms of meat-and-potatoes geometrical measurements, something used to control those small features that aren’t accessible with a touch-probe coordinate measuring machine.

But if you hear the words “state-of-the-art machine vision,” do you think of something different? Production people probably think of their high-speed cameras, feature/character recognition systems, “blob” analysis software and so forth. In fact, if you’re involved in production, you may skip ahead to this article’s “state of the art” subheading. For us metrologists, let’s cover some painful facts concerning our machine vision roots.

There is no denying that the vision inspection system evolved from the combination of microscopes, cameras and optical comparators, to describe it simply. Originally, metrology lab applications were designed to measure two-dimensional part geometry, while machine vision production applications were expected to perform basic pass/fail analysis.

In short, early prototypes of machine vision systems were basically glorified, slightly more intelligent optical comparators.

The dawn of machine vision

Production labs industrywide needed systems that could keep up with production capabilities. They pushed the boundaries of machine vision by demanding more from their vision processors, cameras, optics, computers and software.

Almost simultaneous to the evolution of processor-chip technology, the tasking capabilities of production machine vision systems expanded. So-called “machine vision” soon assumed an ever-increasing role for detecting defects, presence or absence, basic geometrical measurements, part sizing and/or sorting, color recognition and online statistical process control. As production needs increasedand, in fact, are still increasingso did the sophistication of the vision systems and software.

Yet in metrology labs, change came much slower. Vision systems for metrology purposes languished in a time warp, advancing at a very slow pace and, in effect, leaving metrology departments with only slightly more intelligent optical comparators.

Consequently, the metrological world accepted the faults of vision and continued to measure incorrectly. The evolutionary changes occurring in production were not incorporated into metrology inspection systems.

Metrological complacency

Even today, metrology labs often are left with less-than-advanced machine vision metrology tools. The vast majority of vision inspection systems available to the metrologist are only capable of two-dimensional measurements, even though the parts to be measured are three-dimensional. Worse yet, these labs have discarded part datums and are making do with measurements that don’t correlate to these actual reference points.

Metrology labs have accepted the fact that their vision systems measure in two dimensions and that they don’t measure to the datums. Measurement inaccuracies have been accepted by metrologists, who have compensated by developing an excuse mentality for their vision systems. Such an attitude is dangerous and, quite frankly, inexplicable.

Strangely, while metrology labs pushed the limits of touch-probe CMMs, leading to accelerated development for those systems, they didn't push vision systems toward further development.

However, the production world didn’t share this complacency. Production has continued to push the boundaries of vision technology to develop what they need in order to get their job done. Vision systems in the metrology lab have not seen this rapid state-of-the-art growth because metrologists have not demanded comparable improvements to their vision systems.

Reverse migration

Unlike the touch-probe coordinate measuring machine which migrated from the metrology lab onto the production floor, evolutionary advancements in vision technology are migrating from the production floor into metrology labs. Fortunately, the complacency that has simmered for so long in metrology labs is slowly changing with the development of newer systems that incorporate technology learned in the production environment.

The benefits of improved vision systems for both metrology and production groups include improved capability, increased measuring accuracy, decreased inspection times and commonality of process data.

Faster systems

The need for high-speed, accurate and quantifiable vision-based measurements to help decision making is obvious. Compared with other technologies, vision machines are the only tools that can keep up with production and/or measurement of certain required features.

Vision systems’ speed of processing and data acquisitions can’t be matched by any other technology. Thousands of data points in the form of pixels can be analyzed in a fraction of a second for decision making. This is true for production departments and metrology labs alike.

But both these sectors want faster, more accurate systems. The common constraint between production and metrology at this time is that they both measure two-dimensionally. Due to processing time and setup, this is still acceptable for production departments. For the metrological world, however, this is no longer acceptable. Measurements must be three-dimensional and relative to part datums.

At this point, an attainable goal is to accomplish this while spending less money during the process. As a comparison, prices for touch-probe coordinate measuring machine systems have decreased over the years due to improved technology, competition and standardization. Vision systems are just now entering this phase: Yesterday’s two-dimensional vision system, for example, cost the same as today’s three-dimensional system.

State-of-the-art vision systems

For the production sector, state-of-the-art machine vision systems incorporate fast, high-resolution CCD cameras coupled with intelligent frame grabbers that allow multiple-image processing. These advanced systems combine with optics to allow the finest resolution possible.

Image resolution, not necessarily magnification, is the main factor in measuring accuracy. The video-image processor and driver software are key elements in how much can be analyzed and how fast.

As production machine vision roles change, so must the systems. Firmware and software capabilities are changing the quickest. A future hardware change will incorporate high-speed auto focusing to increase capabilities and accuracy even more. For now, however, two-dimensional stand-off measurements for production are the norm.

For the metrological world, state-of-the-art vision systems include those that measure in 3-D and offer powerful analytical processors and software.

Metrology labs don’t require the same throughput as production, although they do require more capability. Therefore, metrology must have the tools to measure whatever production can and at a significantly higher degree of accuracy.

Ultimately, metrology requires the tools to thoroughly analyze production parts or prototypes in full 3-D and relative to part datums. If this is not possible, then the production measurements have no real basis against which to be compared.

As sensors technology continues to push forward, the use of multiple data-acquisition devices will become more commonplace. Production departments will incorporate scanning lasers with vision and/or mechanical contact. The intelligence and flexibility of the vision processors will make possible a handoff of the required information to the alternate sensor for it to perform its functions, thus taking production one step closer to 3-D measurements.

This technology already exists for metrology labs, which combine noncontact 3-D measurements with contact measurements to ensure all measurements are in reference to part datums.

But the proof of any technological advancement is in its usage. The world’s largest supplier of machine tool inserts currently is using on the shop floor machine vision systems with automatic feature recognition. They offer 0.8 micron-per-pixel optical resolution, high-speed digital scanning and automatic geometry measurements to control production of cutter inserts in a grinding environment.

As process geometry changes, the vision system automatically feeds the correction data to the grinders for real-time part geometry correction. In this company’s metrology lab, duplication as well as excess inspection capability is in place to set and correlate each online inspection system.

For this example, unlike many other manufacturing environments, if you were to ask the production and metrology groups what state-of-the-art machine vision is, they both would have an answer.

About the author

Michael Neeves is president of Mycrona Inc., located in Plymouth, Michigan. Mycrona manufactures high-precision multisensor inspection systems utilizing vision, laser and touch probes. For more information, call (734) 453-5880.


[Home Page] [Current] [ISO 9000 Database ] [Daily News] [Phil’s Journal]
[Quality Management] ['98 Past Issues] [Resources] [Advertising]
[Subscribe] [Guestbook]

Copyright 1998 QCI International. All rights reserved. Quality Digest can be reached by phone at (530) 893-4095.

Please contact our Webmaster with questions or comments.

ISO9000 ISO 9000 TQM management quality QC QA teams QS9000 QS-9000 quality digest juran deming baldrige ISO9000 ISO 9000 TQM management quality QC QA teams QS9000 QS-9000 quality digest juran deming baldrige ISO9000 ISO 9000 TQM management quality QC QA teams QS9000 QS-9000 quality digest juran deming baldrige ISO9000 ISO 9000 TQM management quality QC QA teams QS9000 QS-9000 quality digest juran deming baldrige ISO9000 ISO 9000 TQM management quality QC QA teams QS9000 QS-9000 quality digest juran deming baldrige


e-mail Quality Digest