Camera interface standards

It's my impression that the machine vision industry has pretty much standardized on one interface. It’s GigE for area or matrix cameras, leaving USB3 to the scientific/medical community and falling back on CameraLink for linescan applications. (Though I notice Dalsa now has a family of GigE linescan cameras.) However, I know other industries like other formats.

So, when I saw a post on the excellent Adimec blog asking, “Which digital video interface is best for global security systems?” I didn’t expect to learn much. But there were a couple of interesting snippets.

First, regarding GigE, “Processing required to pack and unpack video generates additional heat and uncertain latency…” Now that is news to me. Yes I have noticed a couple of my favorite GigE cameras seem to run very hot, but I hadn’t compared them with USB3 equivalents. Now I think I will.

Second, someone seems to have a bit of a downer on USB3:

“Cons

• Large connector and interface driver
• Maximum throughput unpredictable (chipset, PC motherboard and driver dependent)
• Sustainable speed is much lower than theoretical limit
• Unreliable operation with longer cables (>3 m)”

Interesting points. There’s been so much hype over USB3 that the downsides seem to have been forgotten. Good to see Adimex removing the rose-tinted specs.

This is why it’s important to keep reading the machine vision blogs. You never know quite what you’ll learn. (And kudos to Adimec for providing consistently good content.)

Economic indicators for the machine vision world

It’s third quarter reporting time here in the US, which means I like to see how the leading machine vision companies are faring. Sadly, Cognex is the only publicly-quoted pure vision company: for the others we have to extract what little is made known in their reports and investor conferences.

Two companies with a foot in the machine vision world reported recently: Teledyne and FLIR Systems.

On Teledyne Dalsa…

Using the “Seeking Alpha” financial blog, I learnt that Teledyne are pleased with how “Sales of sensors and cameras for machine vision and medical application increased nicely compared to last year as did sales of infrared sensors and cameras for commercial and scientific applications.”

Answering questions from the investor community, Chair, President and CEO Robert Mehrabian also said, “We think Digital Imaging will grow maybe 3% to 5% in Q4 versus the prior year and Q3. Most of that should come in our DALSA visible camera businesses. We have -- we're doing reasonably well in flat-panel display inspections. As you may know, we make somewhere between 60,000 and 70,000 cameras in DALSA annually, and that business has picked up, especially in Asia Pacific.”

So that sounds pretty positive.

FLIR Systems

This is the other company I’ve started watching. Though not strictly in the machine vision world, they do make and sell sophisticated imaging equipment that gets used in industrial inspection applications.

They’re pretty down about reduced government spending, but CEO Andy Teich did say, “The Commercial Systems division increased revenue 17% versus the third quarter of 2012, with all 3 geographic regions posting positive growth for the first time in 2 years.”

Commercial Systems Division President Tom Surran added, “The thermography line of business grew revenue for the first time since Q1 2012 with all 3 geographic regions posting positive growth.” He went on to comment, “Excluding orders received from government-funded OEM customers, Commercial Systems bookings increased 8% organically year-over-year.”

Looking to the fourth quarter, Tom also said regarding their thermography business, “We will have a strong fourth quarter relative to the quarters so far this year.”

So what’s Brian’s take on all this? Basically, the machine vision business is healthy and growing at a steady click. It’s a good time to be in machine vision.

Renewed R&D effort at Dalsa?

Now that camera-maker Dalsa is part of Teledyne it’s hard to get a clear view of what’s happening there, but the recently published Q2 2013 results and associated conference call offered a few clues. (www.seekingalpha.com is a great resource if you’re interested in conference call transcripts.)

The results showed that while revenues in the Digital Imaging group were off 6% in the second quarter income was up 5.3%. That says to me there’s been some cost-cutting going on, and indeed, as much was admitted during the conference call. It was however also mentioned that industrial machine vision sales were up slightly so the shortfall is elsewhere.

Also in the conference call was some discussion of Teledyne’s acquisition of a company called Axiom IC. Axiom designs CMOS chips, and I’d like to share with you a couple of quotes from the conference call.

First off, CEO Robert Mehrabian said, “This quarter, we made a small but important acquisition for Teledyne DALSA. We acquired Axiom IC, a designer of high-performance CMOS mixed signal integrated circuits. The technologies at Axiom will help us continue developing highly differentiated CMOS imaging sensors and cameras for our machine vision market.”

Then, answering a question about the integration of Axiom, he commented, “…right now, our focus is to get our CMOS imaging development programs improved.”

What can we take away from this?

First, it says to me that Teledyne is serious about investing in the Dalsa camera business. And second, perhaps Teledyne’s management thought Dalsa’s R&D effort needed a bit of a kick in the pants.

And the implications for machine vision cameras?

I think Teledyne and Dalsa are seeing opportunities for differentiated, and presumably higher performance imaging sensors. My guess is that will mean higher frame rates, lower noise levels, and perhaps more development of TDI technology. All this has to be good for us end-users, (and perhaps for Basler too, since there might be less overlap with their business.) The only downside is that these new products will likely command premium prices, but then you get what you pay for.

Color: Why you should stay away, and advice for those who can’t

While writing my previous post, “Color linescan camera options” it struck that more information on color might be useful. In this post I’m going to address why you should steer clear, color camera configurations, and how three chip color cameras work. Now that’s a book-worthy agenda, so mostly I’ll just offer links to those who’ve done the real leg work.

Why you should steer clear

It’s probably natural that the machine vision newbie would think color is superior to monochrome or grayscale imaging. After all, color TV beats black and white. But here’s the difference: you’re not in this for the artistry or to provide entertainment, you’re trying to extract data from a scene. If grayscale is sufficient, that’s all you need.

And to that point, don’t overlook that colored light and filters can let you discriminate between pairs of colors much better than can color imaging. But if you really need to know what colors you’re seeing, or actually want to measure color, then a color camera is the way to go.

If you don’t believe me, read the Dalsa white paper titled “Evaluating Color Inspection: Can Color Machine Vision Improve Results?”

Color camera configurations

Most color area cameras use a Bayer filter over the sensor. Adimec’s blog post “Bayer Area Scan Color Cameras compared to 3-CCD Color Cameras, part 1” (May 14^th, 2013,) explains what this is, and compares it with the three chip camera design.

The Bayer filter design has several disadvantages, but for most of us these are outweighed by its advantage: it’s less expensive than the three-chip design.

How three-chips work

Adimec provide some details in the blog posting, but I found “Separation Prism Technology” on the pages of Alternative Vision had more depth. This shows the various prism configurations, which explains why these cameras are so much more expensive.

So let’s sum up what we’ve learnt:

• Color machine vision isn’t better, just harder
• Bayer filter cameras have their weaknesses but are the right tool for most color machine vision tasks
• Three-chip color cameras use prisms, which makes them good but expensive

Surprisingly enough then, it seems you get what you pay for.

Color linescan camera options

Linescan imaging is complicated, color machine vision is tricky too. Putting the two together sounds like something only the bravest (or most foolhardy) vision engineer should attempt. Yet there are times when it has to be done.

Anyone needing to dip a toe in these murky waters would do well to learn more about the camera options. To that end, Dr. Xing-Fei He of Teledyne Dalsa authored “Trilinear Cameras Offer High-Speed Color Imaging Solutions”, published in the May 2013 edition of Photonics Spectra.

This compares and contrasts the three types of color linescan camera, giving the potential user an understanding of what to look for. Of course, price is left out of the article, but rest assured it correlates with performance, though not necessarily in a linear manner.

Happy (color) scanning!

First quarter 2013 results from Teledyne

Teledyne is the parent of Canadian camera-maker Dalsa, and having last week commented on German competitor Basler’s results it seems only right to do the same for the Canadians.

Unfortunately Teledyne publish far less data than do Basler, but I can share the big numbers: the Digital Imaging segment (Dalsa plus a few other bits,) saw sales increase 8.7% to $102.4M and profit jump 20.9% to $5.2M.

That profit number concerns me, for two diametrically opposite reasons. First, it’s not really a very good margin. Basler seems to have higher margins, although in fairness, I find it very hard to extract the actual numbers from their report. But second, how did they achieve such a jump? Has there been some cost-cutting? Perhaps becoming part of a bigger business meant some overhead could be moved up to head office. I hope it’s that rather than someone taking the knife to R&D.

Time, (and the 2013 final report,) will tell.

Application critique

Including technical details in an application press release lets us all improve our knowledge and advances the machine vision industry. Dalsa, inspect-online.com, and end-user Accuma are to be commended for doing just that with “From Pole To Pole” published April 12^th 2013.

This describes an assembly verification task performed at Accuma where cylindrical posts are measured to ensure the assembly is correct. The solution consisted of a backlight and a Dalsa Boa smart camera.

Two particular issues are described: the need to locate the post in the image, and the importance of ease-of-use and supportability. The first seems to have been addressed by using some sort of fixturing tool to locate the part in the image. (I know that’s obvious to you grizzled vision Pros but it’s not obvious if you’ve never done it before.)

The second issue was addressed by the choice of the Boa camera itself. I have no personal experience but the quote from the Maintenance Supervisor, “"We learned how to program the vision system using the manual that came with the product and we haven't needed any training,"” I found pretty telling. I found the same when I used the latest Keyence vision sensors. (“Easy setup, thanks to Keyence” May 1^st, 2013.)

I would however make a couple of observations on the details of the application.

We’re told an 8mm lens was used. That’s a wide angle of view, so the true diameter of the post isn’t really seen. I would try to go with a longer focal length lens with a narrower viewing angle. Also, given that there’s measurement going on, I would (a) be concerned about the consistency of the working distance from post to lens (if it varies the scale changes,) and (b) want to see some kind of calibration target and process form part of the inspection.

But also, and to show that I can praise as well as criticize, I note that the backlight is some distance away from the post being measured. That’s a good thing because it reduces reflections off the cylindrical post and has the effect of making the lighting “pseudo-collimated”. (Further away would be even better.) Any time you’re backlighting, try to do the same.

So, a good article about a good application, and I’m sure there’s something we can all learn.

Automating inspection in China

Over and over I’m told that “my” vision systems have to be justified with labor savings. I’ve always thought that a simplistic approach because, as we all know, the benefits of using machine vision for automated inspection are many and varied. But that’s how the money people view the world where I work, and that’s why I’m always being told there is no need for my skills in China: it’s cheaper to pay people to look at things.

According to “Machine vision boosts productivity on bottling line,” published March 6^th, 2013 on MachineDesign.com, that’s also the case at beer bottling plants in China where, “Although manual inspection is labor intensive and can cause eyestrain and fatigue, most bottling lines running in Chinese plants operate this way.”

The article also mentions that, “95% of all beer bottles are returned for recycling,” in China. Remind me to stick to canned drinks next time I visit, because I don’t feel comfortable trusting my health to human inspectors. No matter how vigilant they may be, when the glass is flying past at 18,000 bottles per hour mistakes are going to be made.

Fortunately though, as described in the MachineDesign article, Chinese machine builder MingJia has developed automated bottle inspection systems. The article gives details of the system used, (it’s from Dalsa,) but there are no photos. So I turned to the web and found this page about the glass bottle inspector on the MingJia website.

We’re still left with the question of how the bottling plants justify this inspection equipment. I imagine there are two things going on. First, a system built in China is almost certainly cheaper than one from the US or Germany. And second, perhaps the more enlightened beer-bottlers realize that with human inspection the probability of filling and shipping a contaminated bottle is decidedly non-zero. I don’t know about consumer protection laws in China, but I’m pretty sure such a thing would be a big turn-off for customers.

So to wrap up this rambling post: automating inspection should be about more than saving labor costs. Protecting the customer, and thus the manufacturer, should be the priority. Now how about putting a box for that on my capital appropriation form?

Financial reporting season

The few publicly listed US machine vision companies have started publishing their 2012 results, and the Europeans follow in a month’s time, so as the next few weeks roll by I’ll be taking a look at what kind of year these companies had.

We start today with Teledyne, parent of camera-maker Dalsa. Since 2010 Dalsa has been wrapped in to Teledyne’s Digital Imaging group, which includes some smaller companies making IR sensors and LIDAR systems. Overall, Teledyne’s 2012 results tell us little about the machine vision business, because Dalsa is only about 10% of total revenues, but we can extract a few numbers.

First off, the fourth quarter was not a happy one, with a $10.2m increase in Digital Imaging revenues being attributed to a $12.8m gain attributed to newly-acquired Optech. In other words, when comparing apples-to-apples, sales were slightly down.

That’s consistent with information put out by the AIA and others suggesting buyers held back at the end of 2012. What’s interesting though is the growth in Digital Imaging profit, which more than doubled in the fourth quarter versus the same period in ’11. I think this tells us there’s been some cost-cutting going on. Problem is, we don’t know where.

And the outlook for 2013? Assuming those sales deferred from Q4 ’12 come in during Q1 ’13, I think this promises to be a good year. The bigger question might be whether Dalsa’s success can be sustained in the face of continued cost-cutting. I shall be watching Teledyne’s stock (NYSE:TDY) closely, but in the long run, might Basler, as the only other volume linescan camera manufacturer, be the better investment?

Certified System Integrators

I’m very enthusiastic about the AIA’s certification scheme. If you read my comment on the Teledyne Dalsa imaging blog you’ll learn why, but I’ll repeat the key point here anyway: anything that improves the credibility of the vision integration industry is a good thing, and Certification should help would-be integration buyers separate the chiefs from the charlatans.

If you’re interested in seeing which companies have been certified, that information is on the AIA’s Certified System Integrator page. As of writing, there are just five companies, four in North America and one in Brazil, but I would expect to see that number grow rapidly.

One point to mention on this: if you think Certification is simply a case of writing the AIA a check, think again. They actually have some stringent requirements, as listed on that page I linked to.

How to invest in machine vision

If you read “Analyzing the global machine vision business” you’ll know machine vision has excellent growth potential. Today let’s talk briefly about how to seize this opportunity.

I know of only three pure machine vision companies that are publicly quoted, Cognex in the US and Basler and Augusta in Germany. (Augusta is the parent of camera-maker AVT and is pretty much just in to cameras.) If you’re interested in other vision companies you’ll need to investigate their parentage.

For example, Dalsa is part of Teledyne, a far bigger data acquisition company. Likewise, PPT is owned by Datalogix, and Microscan by Spectris. Then there are the sensor companies like SICK and Keyence, which I’m not sure how you’d get into. Investing in any of these means exposing yourself to risk beyond the pure machine vision industry, (which may or may not be a good thing.)

Many other vision companies – probably the majority - like MVTec (Halcon software vendor,) are privately held. If you want a piece of their action you’ll need to talk with the existing owners. Good luck with that!

So are there any other options? Well I think you’re left with starting your own machine vision business, developing some innovative technology, and selling out to one of the big vendors. But here’s a question to my community of readers: are there any private equity or venture capital businesses specializing in machine vision?

An Atom-based vision PC? Really?

The Intel Atom processor is a neat little device. It’s miserly in its use of electricity, so in laptops and netbooks it gives great battery life. It also generates relatively little heat, which makes it good for fanless applications. I’ve seen some camera companies use it in their smart cameras too, but I have to question the wisdom of doing so.

I’ve nothing against the little processor that could – in fact I have one in the netbook I’m typing this on – but they’re not really up to high-speed image processing.

So you’ll understand my surprise in learning that Dalsa – sorry, Teledyne Dalsa – have launched a vision PC built around the little fellow. The box I’m referring to is the GEVA-300. Now I try to avoid being negative, but really Dalsa, an Atom-based machine? And it’s intended for multiple cameras. Why not just step up to an i5? Was the whole point just to have a fanless offering?

And to that point, if you spend just a couple of minutes on the Dalsa site you’ll see they offer such a device, the GEVA-1000. It’s not clear what processor it actually employs, but I’m figuring that the bigger model number means more horsepower under the hood. I don’t know what the price difference is either, but I’ll wager anyone who buys the 300 will later wish they’d stepped up to the 1000.

Financial performance of the machine vision industry

Since most vision companies are privately held it’s hard to glean much information about profitability and growth, though it’s useful to pick out the general trends. For most companies, first half reporting season is upon us, so over the next couple of weeks I intend to share and comment on whatever financial news I can pick up.

First off is Datalogic, parent of PPTVision. In a brief press release issued July 16^th they comment that “the Industrial Automation Division … registered like-for-like growth of over 8%.” Industrial Automation is the division that includes both PPT and Accu-Sort, although as those acquisitions are still being digested it’s a little early to draw any firm conclusions. Nevertheless, 8% sounds good.

July 26^th was when Teledyne Dalsa reported their second quarter results. Dalsa is wrapped up in their Digital Imaging group, which reported revenue of $110.0m, up 15% from Q2 ’11. However, profit dipped 7.9% to $28m. If there was an explanation of this, I didn’t pick up what it was.

Conclusions?

Sales continue to grow, though more modestly than when we were in the rebound years. As for profitability, well I think we’ll have to wait for more detailed information for that. Cognex report next week, Basler and Augusta Technologie (parent of camera-maker AVT) the week later. So check back soon for my perspective on industry trends.

What can TDI do for you?

The TDI in question refers to the Time Delay Integration linescan cameras that Dalsa launched a few years ago. Intended to overcome linescan’s hunger for light, these have a multirow sensor that in effect tracks a region of a web. This gives the effect of a longer exposure time, so letting you run the surface to be inspected at higher linear speeds.

Now realizing that my opening paragraph may have done more to confuse than educate, let me steer you to an article that explains the benefits of TDI cameras by way of a real-world application. “Automating Solar Cell Inspection,” published February 13^th 2012 on the SolarNovus website, is an article from Teledyne Dalsa describing the use of TDI in PV inspection.

This article is interesting for a second reason too. One of the more successful techniques in PV panel inspection is to use IR backlight. Basically, the light shines through the silicon, highlighting cracks and other defects. Unfortunately though, camera sensors have fairly low sensitivity to the near IR, forcing the use of longer exposure times. So with the TDI camera discussed in the linked article Dalsa have increased the IR sensitivity. I think that’s what people call hitting two birds with one stone.

Now back to my question: what can TDI do for you?

Vision-based tool protection

I don’t want you getting the idea that I’m secretly working for Teledyne Dalsa, but here’s another interesting story about an application of their BOA smart cameras. It comes from the February 2012 pages of Control design magazine and is titled, “Avoid a Pressing Problem”.

Most smart camera applications seem to be for product quality control, but in this case the goal was to protect expensive tooling and prevent unplanned downtime. You can read the article for the details; my goal here is to open your eyes to a whole new class of vision-based cost reduction opportunities.

With mold tooling costing tens of thousands of dollars and equipment utilization paramount in hitting cost targets, I’m surprised more effort isn’t put in to these types of applications. Of course, it’s often very difficult to get a good view of the tooling, and the environment can be pretty camera-unfriendly, but I’ve long wondered about some kind of scanner – one of those CIS-type sensors perhaps – that would check everything is shipshape with the tooling between cycles.

This BOA system doesn’t get that fancy, but it shows it can be done.

More smart camera options

If like me, you’re not feeling the love from the nice people in Natick, you could do a lot worse than explore what Dalsa – sorry, Teledyne Dalsa – has to offer.

The BOA smart cameras have been around for a while, and although I have no hands-on experience, they appear a pretty competitive offering. But I noticed earlier this week that there is now a BOA camera that will run Sherlock.

This is a big deal. Sherlock is one of those machine vision software products that many of us cut our teeth on. It’s been around forever, it’s straightforward to use, and it works. So bundling it with the BOA Pro smart camera should make for a very attractive product.

Who needs to feel the love now?

More smart camera options

If like me, you’re not feeling the love from the nice people in Natick, you could do a lot worse than explore what Dalsa – sorry, Teledyne Dalsa – has to offer.

The BOA smart cameras have been around for a while, and although I have no hands-on experience, they appear a pretty competitive offering. But I noticed earlier this week that there is now a BOA camera that will run Sherlock.

This is a big deal. Sherlock is one of those machine vision software products that many of us cut our teeth on. It’s been around forever, it’s straightforward to use, and it works. So bundling it with the BOA Pro smart camera should make for a very attractive product.

Who needs to feel the love now?

Company in the blogosphere

There aren’t that many blogs about machine vision, probably because most people are too busy earning a living to spend time writing it about, so I’m always very pleased to find I have company.

The latest entrant to this niche is Dalsa. Since the start of the year they’ve been putting out the Imaging Blog. So far there have been just five posts, (no, I don’t feel threatened,) but they are interesting articles.

As you might expect from a camera company, it seems as though the blog is going to focus, (how I love that pun,) on the silicon-photon interactions that we call “taking a picture”. And because the people at Dalsa really know how to build a camera, I anticipate some great blog posts with lots of technical detail.

Hope I haven’t raised expectations too high for you guys!

Machine vision acquisitions news

PPT Vision, despite being a pioneer in the machine vision field, has been struggling for a long time. It’s probably good news then that a White Knight, in the form of Datalogic S.p.A. of Italy has ridden to their rescue.

According to the press release on the Datalogic website, they are paying $5.2m for a business with revenues of $6m. That ratio – 86% of revenue – tells us that PPT was not doing well, despite their market reputation, an enviable patent portfolio, and a recently rejuvenated product line-up.

But what does this move tell us about the state of the machine vision industry?

Well to answer that. Let’s take a look first at Datalogic. This is a company with 2010 revenues of €393m (around $515m), making them bigger than Cognex, that is concentrated in code marking and scanning, with related ventures in machine vision (using rebranded products, I believe,) and vision-based retail loss prevention (Evolution Robotics.)

The PPT acquisition comes hot on the heels of Datalogic’s deal to buy code printing and reading company Accu-Sort for $135m (147% of revenues.) (Press release here.) Clearly then, as the Italians say in the PPT announcement, they see “interesting growth prospects” in machine vision.

Implications?

Well, Datalogic are now going head-to-head with Microscan in the code reading and machine vision field. It also means that Cognex is no longer the undisputed heavyweight champion, though their profitability and cash reserves are second-to-none.

Will this prompt a further urge to merge? In the machine vision camera world we’ve already seen AVT parent Augusta snapping up manufacturers with complementary product portfolios, and then there was Dalsa’s acquisition by Teledyne, so I think the answer is yes, expect more mergers and acquisitions.

As for me, I’m going to start a vision company and then sell out to one of these big players.

Application tip for IR lighting

Spotted an interesting little application story on the Vision Systems Designs website a few days back, “Smart cameras solve a pressing problem,” (Andy Wilson, November 1^st, 2011.)

This relates to the use of two Dalsa BOA smart cameras for checking that a die is clear before it closes. As the story relates, closing the die with slivers of stray metal in the way can cost a lot of money.

This seems like a pretty run-of-the-mill machine vision application but there are a couple of interesting aspects. Number one is that it’s the first time I’ve read about the BOA cameras actually being put to work. Number two is the use of IR lighting.

I don’t wish to steal Andy’s thunder, (or his ad revenue,) by giving you all the details here, but trust me, the application needed infrared illumination.

Now here are two tips for when you use IR. First, remember that silicon has low quantum efficiency at wavelengths much over 750nm (and goes to zero about 1,000nm,) so you’re going to need a lot of light to make the application work.

Second, most cameras have an IR cutout filter over the sensor. You’ll need to remove this in order to detect the IR light. This might seem obvious but I have found vision engineers complaining that an IR light isn’t working when all they had to do was remove the filter!

Just a couple of points to bear in mind.