What Exactly is Machine Vision?

What exactly is machine vision and how does it play into today's technology?

IVS Imaging
March 2016 


Cameras are everywhere. Banks, gas stations and almost every other business use them for security, filling DVRs with hours of recordings. People video their cycling adventures, mount cameras on UAVs, and police departments are adopting body cameras as way of documenting confrontations. 

These cameras are all dumb. Even 4k video surveillance relies a human to watch a monitor or recording and decide what's going on. But there is a technology with the ability to automatically analyze pictures and make decisions based on what they contain. This is machine vision.

Decoding bar codes

In machine vision a camera acquires an image. Software interprets that image and makes a decision based on what it saw or found. One of the simplest examples is the modern bar code reader.

Not long ago, bar codes looked like a picket fence; alternating black and white stripes. Today, they're often composed of a matrix of tiny black and white squares. Capture an image of that matrix and software can decode it, revealing a product description, a serial number or even a web address.

Image is everything

Decoding accurately depends on having a well-focused image and plenty of contrast between the black and white squares. Achieving that requires good illumination and a lens that guides the light onto the sensor in the camera.

Illumination is complicated. What works in one situation doesn't necessarily work in another. Much of the skill in developing a machine vision system is in understanding and engineering lighting. Lenses add more complexity, with focal length depending on field-of-view and working distance, and aperture, (f-stop for photographers,), governing exposure time.

Once the light reaches the sensor, (Sony CMOS sensors are a popular choice,) the camera converts it to data. Those bytes are sent to a processor for analysis. Some cameras have that processor built-in, (so-called "smart cameras") but others, such as USB 3.0 cameras, are connected to a PC that does the work.

Machine vision in industry

Machine vision systems are in almost every modern factory. They're measuring parts, checking completeness of assemblies, searching for defects on surfaces and yes, reading bar codes. They replace less reliable methods, helping improve quality and reduce waste.

One of the newest areas is 3D vision. A single camera produces only a two-dimensional image with no depth information. But add a second camera for stereo imaging or project a pattern of lines over a surface, and software can extract heights or distances. This is proving extremely useful in robotics, where 3D vision helps locate and place objects. Vision-guided robots are stacking pallets, picking packages from fast-moving conveyors and assembling devices too intricate for the human hand.

Machine vision elsewhere

Similar software is finding its way into other fields. Cameras read vehicle license plates to catch toll-dodgers, and increasingly, form the core of sophisticated Intelligent Traffic Systems, watching for queues and adjusting traffic lights to suit.

Biometric identification is another application. No longer science-fiction, machine vision systems are performing iris recognition, fingerprint recognition and even picking out faces in crowds. In parallel, growing numbers of research labs are using the technology. Cameras are tracking plant growth, monitoring crop yields and observing the performance of new drugs in laboratory trials.

A technology on the rise

Imaging technology is everywhere. It's in the pharmacy, police cameras and traffic cameras — to give just a few examples. Machine vision takes those cameras and adds automated interpretation. That's valuable in automated industrial systems, and it's finding more applications outside the factory. Pretty soon machine vision will be everywhere! 




Sources:

 http://www.ivsimaging.com/

 http://www.vision-systems.com/articles/print/volume-18/issue-2/departments/leading-edge-views/explore-the-fundamentals-of-machine-vision-part-i.html
Comments