It was my privilege to be around for almost the entire first generation of video measurement, vision measurement or non contact measurement, depending on which nomenclature you prefer. I came out of a conventional metrology laboratory background into RAM Optical Instrumentation in 1993, about a year after the first machines with edge detection were sold. RAM was purchased in 1995 by Newport Corporation and I lasted until 1999. A year later what was left of RAM Optical was merged with Newport's CEJ division.

Gage blocks and micrometers are still around more than 130 years after they first appeared. On the other hand, I believe that about 1/3 of the vision machines made in the 1990-2000 time frame are already retired. Only a small fraction of the 2,000 +/- quantitative machines made by RAM were still getting factory maintenance and calibration as I left in 1999.

In my experience, direct vision measurement performed the usual, simple way can hold 4 microns (0.00016 inches) per inch total uncertainty in the X-Y plane under most conditions, including lack of a calibration.

Vision/video machines in good condition are generally about 20 times faster than similar touch probe CMM equipment and do a much better job of measuring form tolerances in the X-Y plane. Of course, vision systems can measure flat objects like printed circuit traces, soft rubber/foam parts and very small parts that touch probes can't. Using special techniques, the uncertainty can be reduced to less than 1 micron per inch.

Working with parts that aren't flat introduces other errors and slows the vision measurement processes down.

Nearly all user interfaces display 0.1 microns (4 u in) or less. I have not worked with every kind of machine there is, but the best ones I have seen achieve 0.5 microns per inch X-Y precision unconditionally.

Claims of greater performance than these require careful substantiation, in my opinion.

The major sources of error are:
The scales used to measure position and apply corrections,
Illumination issues relating to bulb aging, physical alignment and filament warm-up in high speed lighting transitions,
The unmappable portion of the Abbe' offset error in programmable part coordinate system machines (X-Z and Y-Z,)
Depth of focus present in all optical systems,
Part characteristics like dirt/burrs/chips, inconsistent color/appearance on the part or from part to part, rounded corners/fillets and rough textures,
Motion control system deadbands, including the power zoom magnification control,
The digitizer board in analog systems or the CCD array size in pure digital systems,
The CCD to analog NTSC to digitizer board waveform conversions in all analog systems,
The camera held out-of-square with the stage,
Small angle software errors caused by trigonometry calls and
Software errors caused by rounding and truncation of double precision values.

The programmers I worked with were slow to understand how critical it is that a 1 inch measurement with a +/- 0.25 inch tolerance is "in" if the measurement is 1.250 inches. On the other hand, most customers didn't appreciate that sometimes the result of the calculation was 1.250000000002847e+00 or "out" if truncation is not applied. I waltzed with this issue for years. Hundreds of lines of code were devoted to just this condition and its many permutations. This issue is definitely not at the bottom of the list at 100% inspection and SPC shops.

All of the above holds true for general purpose measuring machines that employ edge detection. Machine vision pattern recognition systems are different. In my experience with this equipment separate artifact parts have to be made that are "too small," "too large," "too far apart" or "too irregular" Then you show each of them to the system and tell it "this one is too small," for example.

I have been looking for a forum where these issues are understood and appreciated. I worked in customer applications about 1/3 of the time while I was at RAM. As an employee, I was bound to follow the directions of my management and limit the scope of my comments. I don't work for any of the more than 100 first generation vision/video measuring equipment manufacturers any more so I can be more open.

I am looking forward to exchanging information and opinions on this subject with others. What I would like to do is get involved with the inevitable next generation of video measurement.