Quality Magazine

Maturing Machine Vision

December 22, 2009
Machine vision quality control technology can be applied to a growing list of applications.

Machine vision has become, in many ways, a mature technology. The key algorithms-pattern recognition, optical character recognition (OCR), edge-finding and geometric analysis, and calibration/gaging-are really about as good as they’re going to get until we start building truly cognitive computers that work like human brains. And I’m not holding my breath on that one happening anytime soon.

But it is always refreshing to find a new application space in which to apply the old tools. This provides an opportunity to learn new concepts and business issues, and then to creatively apply old ideas to the new problems.

Such a space is Intelligent Transportation Systems (ITS). Here, we perform automated inspection of vehicles and their contents. All transportation modes are expanding their use of automated quality control (QC), and while the customers and applications here are not standard fare, the world of quality control applies to cars, trucks and boxcars coming down lanes in many of the same ways it does to widgets coming down conveyor lines. From a machine vision perspective, this all shakes out into several categories.


In most transportation applications, simply acquiring the images can be a challenge. Illumination is often only partially controlled, if at all. The lighting variation from dead-of-night to full daylight often requires the complete gamut of dynamic range control: electronic gain, electronic shutter, motorized iris lens and motorized IR filter to allow switching cameras into pseudo nighttime mode. Basler, for example, has a line of IP security cameras with all of these features, and we have been using them in more and more “machine vision” applications where illumination is hard-to-control. Adding these features into more traditional machine vision cameras, including line scan, will be required for future applications.

Specialty Illumination

LED illumination has become the de facto standard for machine vision applications. Advanced Illumination, for example, has one of the broadest product lines in the industry for both visible and IR illumination using dozens of different illumination techniques. Many of these techniques would be very valuable in ITS, if only we had large enough and powerful enough light heads. These will presumably be developed as ITS evolves.


Figuring out when the part arrives is supposed to be easy. Thousands of mechanical, capacitive, magnetic and optical sensors can be purchased off-the-shelf for a variety of applications. My own personal favorites are through-beam optical sensors since they trip repeatedly and are relatively easy to configure and adjust. But in a transportation environment, running light beams across several lanes is pretty cumbersome, if not downright dangerous.

For this reason, we’ve been bringing back one of my old favorites: video-based triggering. This is a high-end, real-time machine vision technique in which images are acquired at 30 or more frames per second and all images are analyzed for the presence of the vehicle. This requires hard-core machine vision software, but creates a tunable, adjustable and sensitivity-controlled vehicle trigger that is hard to beat.

Velocity Detection

Many transportation applications require vehicle velocity. In law enforcement, traffic flow monitoring and vehicle characterization, velocity measurement is a critical parameter. In ITS applications involving line-scan cameras, velocity profiling is needed to perform image reconstruction and scaling.

Another hard-core machine vision tool comes to the rescue: real-time pattern recognition. By identifying and re-finding interesting patterns in successive images as a vehicle moves by, the velocity of the vehicle may be measured and recorded. This requires real-time image processing, but machine vision technology has evolved to the point where this is fairly standard and can be performed on inexpensive standard PC hardware, if you know all of the software tricks.

The Future

I only have space to barely begin to point out the similarities and crossovers between QC inspection, machine vision and ITS. I hope these ideas will stimulate many brainstorms and product plans providing ways to redeploy machine vision and QC concepts in the fascinating and high-growth area of transportation inspection.