Can machine vision analyze dimples?

According to this interesting article, culled from the Quality Magazine web site (April 30th, 2010) yes it can.

The dimples in question are not those adorning the visage of some gorgeous star or starlet but rather those that are present on earring backs.

Hmm. Not quite so interesting now, but a good application story nevertheless.

Pharmaceutical machine vision application

I love reading application stories: I’m always fascinated to learn how people have solved the problems presented by a specific machine vision task.

One of the problems that I’ve encountered a few times is how to acquire images of long thin parts. If you fit the whole thing in a single field of view you loose resolution and waste a whole bunch of pixels. Line scan seems the obvious way around, but it does require that the part be moved in its lengthwise direction and sometimes that’s just not possible.

PPT has a way around this. They offer an image stitching tool which is ideal for assembling high resolution images of, say, needles. Why do I pick on needles? Well that’s the product covered in “Machine Vision Protects Pharmaceutical Packaging” (Control Engineering, February 2010.)

The actual needle inspection example is on the PPT web site, so you could read the details there too, but stick with the Control Engineering article because there are some other interesting details of pharma applications. One point I didn’t see, which is perhaps the most important reason for using vision in pharmaceutical environments, is that cameras are 100% attentive for 100% of the time. When you cannot afford to ship a single defective item, technology trumps human eyes every time.

The Asparagus Grader never sleeps

In my previous post, “The agri-bots are coming!” I steered you to a story in The Economist about automation in agriculture. Well it’s been pointed out to me that inspection and sorting or grading of fruit and vegetables is already a big machine vision market (actually I have to thank Andy Wilson and a post he made on his machine vision blog for the information.)

Andy was referring to a case study on the Lumenera web site which describes how their USB cameras are being used by a New Zealand-based company, Oraka Technologies, to grade asparagus spears.

The Lumenera write up is interesting, but it’s nowhere near as much fun as watching the video of the asparagus grading machine in action. For that you need to use this link to the Oraka asparagus grading machine.

On a more serious note, this is yet another example of a niche application for machine vision. I’m pretty sure that inspecting asparagus has little in common with sorting potatoes, grading carrots, or counting peas, so with all of these, a unique application has to be developed. Kudos to Oraka for finding an exploiting an application with replication potential!

The agri-bots are coming!

One of the big potential markets for machine vision is agriculture, or so says The Economist in “Fields of automation” (December 10th, 2009.)

The point is that it’s becoming very difficult to find the people to do the back-breaking work of picking fruit, and as plants insist on exhibiting random variations in the way they present their produce, automation is going to depend upon machine vision.

Great news for those of you looking for a big new application area!

Build your own camera?

How many integrators do you know who build their own cameras? Not too many, I’ll wager. So you’ll be as surprised as I was to learn that Key Technology, out of Walla Walla, Washington, build their own trichromatic cameras. These are designed to image in the IR, visible and UV wavelengths, and as Key are in the food inspection business, I’m assuming there’s something about food that requires multispectral imaging.

I have to wonder though, isn’t it easier and cheaper to buy three separate cameras, or are their benefits from having three sensors looking through the same lens?

It was actually a story in Vision Systems Design (August 27, 2009) that led me to the Key web site. This concerned a potato inspection system developed by Key that they’ve christened Manta®. Even if potatoes aren’t your thing, it’s an interesting story.

Machine vision can make a bowl feeder smart

Bowl feeders are a great way to orient parts for assembly, (providing you can justify the tooling expensive a changeover times,) but there are some geometries where it’s not easy to discriminate between similar parts or poses.

Enter the smart bowl feeder!

As reported in Assembly magazine, (“Vision Enables Feeder to Handle Multiple Parts,” (July 27, 2009,) Rixan have added machine vision and a robot to the humble bowl, thus giving it much improving part sorting abilities.

The article, which I encourage you to read, gives quite a lot of detail about how the PatMax tool from Cognex, combined with various lighting configurations, is used to identify parts by their shape. It’s a good application story; take a look.

Machine vision rocks!

Here’s an interesting application story: Prosilica cameras are hard at work in South Africa, measuring the size of rock fragments coming out of the crusher. This blog article, Machine-Vision Rock Solid in Harsh Mining Environments gives details of both the application and the environment. Personally, I can’t think of a worse place to put a vision system than a hot and dusty rock-crushing plant.

The Lynxx Optical Particle Analyzer was developed by a company called “Stone Three,” and if you go to their web site you’ll find some screenshots (follow that link then click the thumbnails on the right of your screen,) that illustrate what it does.

Unfortunately no pictures of the camera and light setup, so if anyone from Stone Three is reading this …

Great application story!

Why aren’t you getting coverage like this for your machine vision applications? This is a story about a vision application in the print industry that saves nearly 300 tons of paper a year. That hits both the financial and environmental hot buttons and makes everyone look good.

The integrator responsible for this application is Imaging Technologies, Inc., of San Jose, California. Their web site has some interesting videos but is sorely lacking as to details of the technology they employ. I’m guessing it’s line scan but I’d like to know more.

Machine vision is moving outside the factory

Those looking to grow our industry are starting to look beyond industrial inspection applications. The new frontier, according to Winn Hardin (“Security Comes Knocking on Machine Vision’s Door,” Machine Vision Online, August 11, 2009,) is security applications. It seems an obvious application area: we all know that humans are lousy at monitoring tasks while computers never get bored or distracted. But while CCTV is great for recording events, so far there has been little effort at adding image analysis capabilities to security cameras.

You might wonder why this fusion of image analysis with security hasn’t happened already. Well one of the challenges is that the environment in which security cameras have to operate is far less predictable than that in the factory. Day and night, for example, then there are stray reflections, wet surfaces when it rains and so on. But challenges create opportunities for entrepreneurs, and there’s a lot of work going on.

For a great summary, you might like to read “Smarter Video Analysis Techniques Mine More Data,” (Electronic Design, July 23rd, 2009.) Author Richard Quinnell explores the efforts underway to develop smarter algorithms, incorporate more intelligence into cameras, and gain a better understanding of application issues. For an example of the last point, he notes that it’s usually preferable to work with monochrome images rather than color. Color images are too susceptible to variations in lighting, which is something that can rarely be controlled in security apps.

Bottom line; if you’re looking for new opportunities in machine vision, look to see how the tools we use in manufacturing can be applied to CCTV applications.

Machine vision for machine guidance

If you perform laser cutting or marking in your facility you’ll know that while the machine might be very repeatable, the blanks and the fixturing are less so. This means that the target will be in a slightly different place every time.

One way to deal with this is to spend more money on higher precision blanks, and yet more money on better tooling for repeatable blank location. However, a smarter alternative might be to give the laser “eyes” so it can decide exactly where to put the beam.

That’s what Worldwide Laser of Gilbert, Arizona have done with their laser vision system. Machine vision is used to locate the target edges or region. These coordinates are then input to the machine controller which drives the laser to the required location. Seems like an elegant solution to the problem.

Detecting porosity

If you’re involved in machining cast metal you’ll know what I mean: those little cavities that are revealed only when you cut metal. They can result in a bore not sealing, or in premature part failure, so it’s pretty important to find them before shipping or assembling the machined part. In many factories this detection work is done by a small army of motherly women (why are they always women?) sitting at the end of the machining lines.

Well bad news ladies! Valentine Robotics, a distributor and integrator of the Scorpion Vision software package, have come up with a machine vision method of detecting porosity. This posting on the Scorpion Vision blog gives some details (nothing on the lighting though, which is a pity.) To be perfectly honest, from the pictures provided the image processing task doesn’t look that hard. I feel sure a smart camera could have done the job. On the other hand, a base version of Scorpion is pretty inexpensive.

If you want to know more about Valentine, here’s a link to their web site: www.valentinerobotics.com

Look outside the factory for replication potential

Many users of machine vision – myself included – think in terms of manufacturing applications of machine vision. Unfortunately though, there tends to be limited replication potential for these industrial systems. As a result, the development cost per system is high (perhaps we could name this the “Grey ratio”?)

However, some developers take a broader view, and are finding vision applications that can be copied many times. Back in December 2008 I covered a system marketed at golf clubs and driving ranges – the “Swing & See” – and now here’s another: the robot pharmacist . (This opens up as a pdf.)

These are both great applications, but the automated pharmacy strikes me as absolutely brilliant. This is a product positioned to benefit from aging populations in Japan and the western world, and someone deserves to make a great deal of money by spotting the opportunity.

By the way, I’d like to acknowledge Allied Vision Technology for sending me the newsletter that publicized this application. Use this link – newsletter – to read the whole publication, and be sure to check out the cameras on their web site.

VSD Video Library

You probably believe that it’s possible to learn something from every factory you visit, but who has the time and money for ‘industrial tourism’? Well thanks to Vision Systems Design, it’s no longer necessary to endure airport security in order to see how someone solved an inspection problem. Just dip in to their video library and browse the factory automation, R&D and other offerings.

I particularly enjoyed the ‘Robotic bin-picking of water pump parts,’ if only because, after the robot moves the acquired object off camera, it sounds suspiciously as though it drops it onto a concrete floor.

Machine vision success story

Machine vision is not new, yet I’m still surprised by how many people in manufacturing treat it with a suspicion usually reserved for politicians and emails from Nigeria.

News flash folks!! Machine vision is a proven technology. It works!

That message should be reinforced by this case study culled from Quality Magazine (March 30th, 2009.) In “Quality at 400 Blades per Minute” Managing Editor Maggie McFadden describes a factory that has made vision part of its standard equipment.

The part that really caught my eye was this: “… with 10 million [razor] blades made at the plant daily, it’s impossible to rely on the human eye for inspection.” The article then goes on to list some of the machine vision applications to be seen in the plant.

Notice how people at the plant dismiss human inspection. Although the justification is not spelt out, I think it’s safe to assume that:

(a) An awful lot of inspectors would be needed to 100% inspect 10 million blades per day. And at a fully loaded cost of say $20 per hour, (the plant is in the US,) that would make the product just too expensive for the market, and

(b) If we accept the oft-quoted figure of human inspection being only 80% effective, there would be the potential for a lot of poor quality blades to reach retailer’s shelves (and perhaps my chin.)

So the smart people at the plant just put machine vision to work. And I’m willing to bet that the more systems they install, the easier it becomes. It’s just a learning curve thing; give it a go, you might like it.

Experience is a great teacher

Here’s a novelty: a press release that actually tells us something useful.

“Machine Vision in Wood Industry” tells the story of a vision system developed by an organization called Forintek, which apparently is Canada’s Wood Product Research Institute. (I’m neither Canadian nor a lumberjack, so I’m unfamiliar with the body.)

The system was developed to measure the quantity of “fines” in lumber being used for OSB structural wood panels. However, the goals of the system are not especially relevant. What caught my attention was the description of the vision system development process. Apparently, the first attempt was not too successful, because the developers went back and had another go.

The press release does a nice job of explaining the problems experienced, and why the various upgrades, such as a GigE camera, strobed LED lighting and LabVIEW software, were adopted. Apparently the system now functions much better and is being used by more than 10 OSB mills.

Experience is a great teacher. Take a few minutes to learn from the experience of others.

Medical Machine Vision

A few days ago I expressed concern about the use of machine vision for surveillance. I still maintain that there should be greater public awareness and discussion of this issue, but I also recognize that there can be benefits from surveillance.

Over in the UK, two unrelated Doctor Smiths from the University of West England, (UWE) are working a system to monitor infant breathing in hospitals. As I understand it from the article, their approach is to use structured light and image comparison. The structured light provides the 3D information, which will change slightly as the infant breathes: comparing images will show the breathing rate, and I imagine, even the intensity (short, shallow breaths, or long, deep ones.)

This strikes me as a great project that could yield major benefits for hospitals and their patients. Given the replication potential – x beds multiplied by y hospitals, where both x and y are big numbers – this has some serious commercial potential too. Dr. Smith and Dr. Smith deserve all the success they can reap from their efforts.

Application case studies

It’s very difficult for someone with little machine vision experience to figure out how to approach a new application. Everyone tells you that lighting is key, which of course it is, but unless you have some grasp of the basic image analysis tools available it’s not obvious what you should be trying to achieve with the lighting.

That’s why I was impressed by the machine vision application examples appended to the “Vision Systems Design” article by the Soliton guys. If you scroll to the end you’ll find five examples of real vision applications, with a discussion of the logic used in setting up the inspection. This is what’s missing from so many vision papers and articles, yet it’s exactly what engineers need in order to figure out to approach their particular job.

Great work Soliton, keep it up!

Plenty of time for marketing

Some years ago I ran my own consulting business. The economy was good and it took only a gentle shake of the tree for some nice projects drop into my lap. Unfortunately though, while I was busy doing chargeable work I would omit to continue my marketing efforts, with the rest that my order input had a serious stop-go quality to it.

I share this with you because I’ve noticed in recent months how the trade press is full of articles from machine vision integrators sharing their latest and greatest applications. Cynic that I am, I have to believe this flurry of interesting media stories stems from a lack of chargeable work. In other words, with a lack of new orders coming in, these enterprising businesses decided it was time to get their name out in the public domain.

While this lack of work is obviously bad news, it does have a silver lining for those of us who like to read about real machine vision applications: there are lots of stories being told.

For instance, turn to the “Vision & Sensors” section of Quality Magazine. In the January ’09 issue you’ll find intelligent pieces by David Dechow, “Integration: making it work,” Dan Reed and David Wyatt, “Raising the bar,” (about reading datamatrix codes,) and Maggie McFadden, “What’s the Big Deal with Light Sources?”

The machine vision business may not be booming, but there sure are some interesting stories to read.

Good news on capital spending

Need something to lift your sagging spirits? Well how about this: the lead story in the December ’08 edition of Assembly Magazine discusses the 2009 investment plans of US manufacturing companies. And their conclusion is that, while “rosy” may be too strong a term, things don’t look bad at all.

The article, “A Modest Increase,” uses a host of pie charts and graphs to explain who will be investing, where they are, and what they expect to be buying. Now machine vision is not discussed explicitly; the focus is on assembly automation. But since vision is becoming standard equipment on high-end assembly lines, it seems reasonable to conclude that smart cameras and vision sensors are going to be bundled in to those kinds of purchases.

So if you’re selling machine vision, the magazine article will give you an idea of who you should be talking to. And if you’re searching for a job that will let you utilize your machine vision expertise, again the article tells you where to look.

Bottom line: opportunities abound – go find them!

The machine vision golf pro

We’ve seen machine vision applied to tennis, cricket and soccer. Now, according to Advanced Imaging magazine, it’s the turn of golf.

Systems for recording and analyzing a golfer’s swing have been around for some time. What’s new about the “Swing & See” system from V2S is that it's a self-service tool. In other words, you go into the specially-equipped booth at your local driving range, drop a buck, (maybe more,) into the machine, and a few minutes later you’re watching a video of your swing. What’s more, this is high-speed video taken from two directions, so you can get a close look at what you’re doing wrong. Better still, you can download it onto a USB memory stick so you can show all the guys at work just why you have that wicked slice.

Should your golf pro start thinking about a new career? Maybe not, but it could be a very interesting way for driving ranges to generate some additional revenue.

By-the-way, I couldn’t find a web link for V2S, so if anyone knows where they can be found, please share it here.