Modern vision systems

Features - Vision Systems

What’s ahead for the next generation of vision systems.

April 29, 2021

This is an example of artificial intelligence (AI) training to locate defects on parts.

There are three common types of vision systems: 1) CNC vision, 2) manual vision and 3) large field of view (FOV) vision. However, it’s important to note that there can be overlap across these categories (e.g., a large FOV machine can be a CNC machine or a manual machine).

While many facets of these technologies haven’t changed, during the past five years one development has emerged – stitching. Some vision technologies now generate large FOV images by stitching together multiple views of a part to see a greater overall image rather than importing the FOV all at once. This is a faster way to see and measure the part in total, and it’s also easier on the operator. And because high magnification is required to zoom in on small features, stitching combines speed and magnification to create a larger FOV than originally possible with the vision machine itself.

This is appealing for medical companies, given that large FOV machines on their own can detract from accuracy, making them less ideal where traceability and accuracy are crucial and where measurement is integral for U.S. Food and Drug Administration (FDA) approval.

As an alternative to costly large FOV vision machines with stitching capability, manual vision machines are also an option. Depending on the machine, exceptional optics can still measure the Z-axis with automatic focus – except the operator must manually move the stage.

Next-gen AI software

One exciting development in vision machine measurement is the recent arrival of artificial intelligence (AI)-based software. Used primarily for defect analysis and detection, and paired with a third-party camera, this software is teachable. Give the software good images and tell it why they’re good. Then supply the software with bad images and highlight the defects. Consistently increasing the number of images trains the software to automate defect analysis to separate good parts from bad.

Removing the operator from the analysis results in faster inspection of finished parts. This automatic, expedited defect detection capability is crucial for medical parts from needles and syringes to any part with zero tolerance for scratches, nicks, or molding defects.

Next-gen vision machines

The next generation of machines will be built for even greater measurement speed without sacrificing accuracy. They’ll be able to keep things flowing without interruption. Users will take a rapid succession of measurements because the machine will move across the features without stopping. In the near future, vision machines will allow traditional CNC machines to accelerate throughput with one continuous data stream.

An LED ring light surrounding a camera.

Illuminating the future

Like a microscope, a vision machine requires a light source for proper measurement, and there are different ways to light a part. A stage light can illuminate a part from below to form a shadow, which produces contrast and is ideal for picking up outside edges and through-holes.

However, sometimes sensors measure a notch, a blind hole, or other internal feature. In this instance, illuminate the part from above using a coax light. This type of light shines straight down through the camera lens and uniformly illuminates the surface for measuring surface flatness.

Another near-universal lighting option is a ring light. This also illuminates the part from above, but instead of shining straight down, it shines at an angle. This works well for measuring the diameter of a cylinder. Don’t use a coax light because it will eliminate the top of the cylinder. However, a ring light will bring out its 3D features. More advanced vision machines allow changing the ring light angle, or use only certain quadrants of the light, depending on measuring needs.

Historically, vision machines used halogen lights for stage, coax, and ring lights until about a decade ago when manufacturers moved to LED lights. Before then, LED lights were either not bright enough or too expensive. They are now the lighting standard, because they last longer than halogen lights and their intensity can be set almost instantaneously.

The three main types of light sources are likely to become even more fine-tuned and powerful. Machines in the future will have greater sensitivity and the ability to manipulate lighting angles in a more sophisticated manner.

An example of image stitching where three camera fields of view (FOV) combine to create a single, measurable image.

The future: CMOS

In the past, all vision machines used a charge coupled device (CCD) sensor or global shutter to capture frames and change the focused image into an electric signal the computer can understand. A CCD captures the whole frame simultaneously.

However, CCD is an old technology with limitations. That’s why many current and future vision machines will use a complementary metal oxide semiconductor (CMOS) image sensor instead. CMOS has abundantly more pixels compared to CCD, which increases accuracy. And while previous CMOS sensors would output too much noise to be useful for vision machines, sensors have recently improved signal-to-noise ratio to where they can now be used in vision machines. Sharper images and a higher pixel count increase measurement accuracy, crucial for measuring any products, but especially those in the medical field.

A mix of manual and automatic will remain

Despite innovation and newer advancements in next-generation vision machines, the future will always hold manual vision machines because of their lower cost and familiarity factor.

Every shop has likely used a profile projector, and it’s a small learning curve to go from a profile projector to a manual vision machine. And, for smaller shops, cost is a significant consideration. If a shop can only afford one vision machine, it’s likely to choose a manual machine rather than a CNC machine.

However, for those less constrained by budget, the future will be automatic, with larger FOVs and greater measurement accuracy. That’s a good thing for all fields, including medical.

Mitutoyo America Corp.

About the author: Mark Sawko, vision product specialist at Mitutoyo America Corp. can be reached at 888.648.8869 or