Today, there are many choices of video measuring machines available on the market, each of them designed to suit different applications and achieve different levels of performance. One of the biggest differences between video measuring machines are the video cameras used to collect and analyse the optical image. Written by Tim Sladden, OGP.
What makes a good metrology camera? The first thing to understand in discussion of vision camera technology is that not all video cameras are suitable for measurement. It’s a common misconception that a higher resolution camera (ie: one with more megapixels) is inherently better than a lower resolution model. With 12, 14 or even 20-megapixel cameras available in today’s mobile phones, it’s easy to expect that a precision measurement system should have as many or more pixels.
In fact, cameras intended for precision measurement rely on more important factors for their quality. Very high-resolution cameras come with practical problems that make them less than ideal for measurement.
First, metrology cameras must be stable throughout the range of conditions in which they will need to operate. Thermal stability – making consistent measurements over a range of operating temperatures – is one key characteristic. Not only must the camera remain stable through normal temperature variations, the camera itself must be cool-operating so it does not contribute heat to the optical system. Many high-resolution cameras get quite hot when operating at full frame rates – some becoming literally “too hot to handle” with bare hands. These cameras are not well suited for precision measurement applications.
Second, metrology cameras must have highly symmetrical pixels – pixels must be perfectly square and arranged orthogonally in the array with equal spacing between them. Arrays should have minimum opaque areas – deadwood between pixels that are not light sensitive – so that as much of the array as possible is available to image the part being measured. Only a small class of video cameras have an array of sufficient quality for high accuracy measurements.
Third, the pixel size must also be a good match with the magnification and resolution of the imaging optics, and with the feature sizes to be measured. A high megapixel camera with very small pixels combined with low quality off-the-shelf optics is far less capable than a sub-megapixel camera with a pixel size well-matched to the resolution of the system optics.
Fourth, the supporting electronics of metrology cameras must be very good. Metrology cameras must have very high signal to noise (S/N) ratios to enable accurate measurement with real time image processing. They must be highly linear – meaning that there must be a proportional change in output for each corresponding change in light intensity in the image. And they must have a wide dynamic range – the ability to detect subtle intensity changes in scenes that are either quite dark or brightly illuminated.
And of course, metrology cameras – and their manufacturers – must have high quality and reliability over the long haul.
Analogue vs Digital – Which Is Better?
The terms analogue and digital can be quite confusing when it comes to metrology cameras. Most “analogue” cameras used for measurement today are mostly digital, and most digital cameras are at least partially analogue. How do we make sense of it all?
All video cameras use a sensor comprised of pixels – individual light sensors – that output a minute electrical signal which is proportional to the amount of light they sense. This is inherently an analogue process – with the conversion to “digital” happening as the electrical impulses from each pixel are processed by the camera’s electronics. Most analog cameras rely on a “frame grabber” circuit board that captures, stores and digitises the pixel data from each snapshot. Typically, the frame grabber converts the digital image back to analogue form to display it on the system monitor.
Digital cameras have their A-D conversion electronics on-board, and plug directly into the PC where the pixel data for each frame is processed (digitally) by the camera’s driver software. These two techniques provide the same basic data to the measurement software, but they do it in slightly different ways. So, what makes digital better?
First, digital cameras eliminate the need for a dedicated interface (frame grabber) board in the system computer. This means lower cost, better reliability, and in the long run, greater flexibility for the user because the computer and operating system can be upgraded more easily than one which must operate a particular circuit board.
Second, digital cameras use advanced electronics that enable faster and more intelligent processing of the pixel data. The advantages here are small for traditional VGA or CCIR format cameras with about 300,000 pixels, but as the camera resolution increases to 1.0, 1.5, 2.0, or even 5.0 million pixels, the advantages become clear. Trying to process one million or more pixels using a traditional analog frame grabber with an A-D conversion and a subsequent D-A conversion for the display would simply take too long. Real time imaging would not be practical during automatic measurement.
Third, digital cameras enable on-the-fly image enlarging, also known as digital zoom. Digital zooming is actually a change in the ratio of camera pixels to video monitor pixels. As the “digital zoom” level is increased, the number of monitor pixels used to display each camera pixel is increased, making the image appear bigger on the screen.
Digital zooming does not increase the optical magnification of the image presented to the camera, and thus does little to improve measurement accuracy or resolution. What it does do is make it easier to observe the image, and place video measurement tools at precise locations, which does indeed improve the robustness of measurement routines. An enlarged image is also useful for visual inspection and record keeping.
Keep in mind that field of view size – the area that can be viewed in a single image – is reduced as the optical magnification goes up. The trick is to balance the magnification needed to measure features accurately with the convenience of being able to see and simultaneously measure a larger area.
Digital cameras also enable the enlarged image to be panned – scrolled in any direction – to view a different portion of the original image when it is zoomed up. The combination of panning and zooming enables scenes with small features to be blown up and moved about the screen for easy viewing.
Taking good advantage of high-resolution cameras with digital zooming requires good optical resolution – by good, we mean resolution that is a good match with the pixel size of the camera. A good match is one that typically has 7-10 pixels in transition where an edge occurs in the image. Of all the megapixels in the camera, it is this small handful of pixels in transition that actually make measurements. Consider this example:
As the camera reads out pixels from left to right in this image, the intensity curve would look something like this:
The areas of interest are on the transition from dark to light. If we look closely at a plot of pixel intensity in this transition area, we’d see something like this:
The transition from dark to light spans several pixels. Analysing the rate of change between adjacent pixels is a more precise means of determining the edge location than binary processing because it ultimately allows sub-pixel resolution of each measurement.
If the image is so sharp, or the pixels so large, that the transition occurs in just one or two pixels, the image processing software has very little data to work with, and the subsequent measurement will not be as accurate or repeatable as it could be.
Since industrial metrology covers a broad range of parts, feature sizes and tolerances, manufacturers typically offer several different optical system and camera combinations.
For applications where the feature sizes are large (millimetres), moderate magnification levels are required, and we can enjoy the convenience and flexibility of a zoom optical system. For this type of application, a 1.5 megapixel camera with nominal pixel size of 6-8 microns provides good performance.
For micro fabricated parts with feature sizes less than 100-microns, a fixed lens optical system with a 5X to 100X High N.A. objective lens is typically used. This arrangement will present a sharply focused high-resolution image, and will benefit from a 5.0 megapixel camera with a pixel size of around 2 microns.
In between these two extremes, we see a variety of high resolution zoom and fixed lens optical systems that work well with medium and high-resolution cameras. The key is matching the camera’s pixel size and resolution to the optics, and the magnification to the feature size.
The Future with Digital Cameras
The benefits of moving to all digital cameras are numerous, offering better measurement capabilities, high reliability and lower costs. A properly qualified digital metrology camera will enable faster measurement with better resolution and accuracy than the same system with a traditional analogue camera. The key words here are “properly qualified” – not just any digital camera will do.
More pixels will provide improved resolution, provided the optical system and camera pixel size are a good match with the parts and features to be measured. Digital cameras offer lower cost and greater reliability. There is no video card, therefore a more compact computer can be used. Connections are simple, and there are fewer components that could fail.
All digital image processing is fast – particularly for very high resolution cameras. Digital panning and zooming enables convenient program set up, while enabling programs to run at a lower magnification – with a larger field of view – than would be possible with an ordinary analogue camera. Finally, digital images are easily stored, allowing images to be saved for reference, or even re-measurement off line should the need arise.
CHECK OUT OUR LATEST ISSUE!
WANT MORE INDUSTRY INSIGHTS? SUBSCRIBE TO IAA NOW!