A 4-sensor line scan camera

I hear through the grapevine that JAI are launching a new color line scan camera. Unlike existing models of color line scan, which use either a bayer filter over the CCD or 3 sensors with individual color filters, this uses JAI’s prism technology. The major benefit is the superior color registration that results from a common optical path. This eliminates the alignment problems that make color line scan cameras such a pain to use.

Incidentally, the camera is available in only a 2k format, but the CCD sensors cover Red, Blue, Green and Near IR. This should make it useful for inspecting organic materials where IR can help uncover more data than is evident to the human eye. For an example of where this can be useful, take a look at this page on the JAI site about the similar multi-spectral AD-080 camera.

As for the LQ-200CL, which is what they’re calling this new camera, at the time of writing there are no details on the JAI site, but I’m sure that will change soon. Also absent is any pricing information. It won’t be cheap, but judging by the pricing of the AD-080 range I would expect this to be around $3,500.

Another source for machine vision cables

Cables are the unsung heroes of machine vision: if it weren’t for their sterling efforts all that camera work in converting photons to bits would be for naught. Yet so often the cable is an afterthought. And when we do come to hook camera to PC, how often do we find that the cable on hand wont reach, or the connector doesn’t fit? Add to that the need for screw-in connectors and suddenly you’re scratching your head wondering where to go for parts.

Wonder no more, because today I would like to share with you a new cable vendor. Well they’re new to me but I believe they’ve been in business for some time. I’m talking about Newnex Technology Corp.

California-based Newnex offer an extraordinarily wide range of Firewire, USB and GigE cables, and if you still can’t find what you need they’ll even make up a custom cable for you.

Something to add to your “Favorites” if nothing else.

In praise of high frame rates

I see and hear lots about the value of high sensor resolution, but almost nothing about what high frame rates can do. I think that represents a missed opportunity.

The benefit of acquiring many images, when you’re performing part inspection, is that it provides, as the TV people say, “another chance to see.” Let me explain: imagine a situation where a part moves under a camera. The camera receives a trigger and acquires an image. There’ll be some slight positional variation image to image due to the inherent variation in the system (jitter and so on.) Odds are, this variation also means the lighting varies subtly from image to image.

Usually lighting variation is the enemy, but I’d argue that by capturing multiple images it can be your friend. Say you’re trying to read a pin stamped code on a moving part. Rather than grab a single image and hope for a good read, grab several images and see which one gives the best read.

This is why I’m a fan of the Cognex Checker. Yes, the resolution is low, but the frame rates are very high, so with a little imagination this sensor can be used to perform some quite interesting inspection tasks.

Mobile machine vision lens calculator

You knew it was going to happen, and now it’s here: there’s an iPhone App that lets you work out what lens to buy for your vision application. This comes courtesy of Edmund Optics, who presumably believe that engineers on-the-go will be thrilled by the ability to whip out their cell and select a lens.

As you can probably tell, I’m a touch skeptical. I’m not an iPhone user so I haven’t evaluated the App but I find it hard to believe it can outdo the MachVis program from Linos. For me that neat little program, which requires a PC to run on, is the gold standard of lens calculators, and it will take more than an Edmund Optics sales aid to budge me.

Incidentally, I should thank Andy Wilson for alerting me to this new. Here’s a link to his blog posting, “Apple app sorts lens configurations.”

Cognex ratchets up the resolution

I’m a little surprised how long it took, but Cognex finally got around to releasing a 5 megapixel version of their In-Sight smart camera, the 5605. Specs are sketchy, to say the list. There’s a single sentence on the In-Sight product page, but no indication of frame rate or the processor employed.

For me those two factors are critical in deciding whether to actually deploy vision applications on the 5605. All the resolution in the world is no good to me if I can’t capture enough images or can’t finish processing one before the next arrives.

But I suspect those points will be moot when I find out the price. I doubt this little baby is much below $9K, which means that I could save a pile by going with a PC-based solution, perhaps using a camera like the Pike from AVT, (which I’m guessing has the same Sony ICX625 sensor as the 5605,) coupled with Cognex VisionPro.

Image sensor projections signpost machine vision opportunities

Every machine vision system uses one or more imaging chips, be they of CCD or CMOS construction. So it would seem reasonable that we can piggyback on the market projections of the sensor manufacturers to see where the growth will be.

“PC’s Semiconductors Blog,” (June 22nd, 2010,) sets out some interesting, and perhaps rather bullish, projections for growth. What really caught my eye is how they anticipate an explosion in automotive imaging applications over the next few years. This would seem to add weight to my predictions for machine vision in the automotive world – once those images are available doesn’t it make sense to add automated analysis? If not, are we expecting drivers to add the monitoring of display screens to their list of duties?

Automotive machine vision – it’s more than a business opportunity, it’s your moral and ethical duty!

Simplifying machine vision integration

Everyone talks about easier application development but what about easier integration? In my experience, actually getting a vision system to work can be every bit as challenging as putting together a robust inspection algorithm yet it seems to be an area that’s fallen through the cracks: neither the camera vendors nor the software developers want to get into it.

Until now. I’ve been reading about the “GigE Trigger Device” from Baumer. This sets out to replace the PLC that’s conventionally used in most less complex vision applications. I haven’t actually used the device yet but from reading the literature it seems it’s designed to allow encoders and proximity sensors to be wired in directly. This enables it to send triggers to the camera as well as signally other actuators, such as an air blowoff, I suppose.

If it works as advertised this could be quite a significant advance in integration technology.

As a side comment, I should mention that Gardasoft have a somewhat similar product; their CC32 Trigger Timing Controller appears to perform the same I/O functions as the GigE Trigger Device although without being optimized for Gigabit Ethernet.

That’s two devices intended to make integration a little easier. I’m sure others will follow.

Getting started in machine vision

Ten years ago I used to see a lot of “How to do your first machine vision project” types of articles in magazines and on web sites, but these days they seem rather few and far between. I suppose that’s a sign of how the machine vision industry is maturing: it’s becoming just another tool rather than a “when all else fails” last resort. That said, I have to believe there are still engineers out there about to undertake their first vision project.

For such “newbies” Matrox have posted a paper, “How to Plan Your First Vision System” on the AIA web site. It’s not a bad effort though for my money it tries to cover too much ground in too short a space.

A better start point might be the free “Fundamentals of Machine Vision” course that Perry West provides on the web site of his consulting business Automated Vision Systems, Inc. You will need to register, and it is of course a hook to lure you in to their consulting services, but I’m sure you’re smart enough to understand that.

It might also be worth me explaining what makes Perry such a good source. Perry has been working with machine vision since the dawning of the industry. He understands the fundamentals and he knows how to apply the technology. In short Perry West is machine vision, so his course is worth taking.

Machine vision brain snacks updated

Some years back I stumbled across a neat little feature on the Coreco web site called, if memory serves, “Bob’s Brain Snacks.” These were short, blog-style postings in which “Bob” explained in simple terms how various image processing tools actually work. I found them a very useful primer.

Well the “Brain Snacks” have gone the way of Coreco, but I noticed recently that Matrox are offering something similar. Under the heading of “Vision Squad Files,” Matrox provide a series of tips, tricks and explanations. They’re not quite as to-the-point as “Bob’s Brain Snacks” of yore, but they do a good job of explaining how to use various tools to actually get something useful done.

Use the link I’ve provided to reach them as they’re rather hidden away on the large Matrox web site.

Alternatives to machine vision

While my main interest is in using machine vision to help achieve business goals, it’s worth noting that it isn’t the best tool in every situation. Take crack detection for example.

It can be very difficult to find a crack in an image: lighting is hard enough, but the biggest problem, in my experience, is that while a crack might be quite long it has almost no width. This means it may be less than a pixel wide, which means it can’t be detected reliably.

When you’re looking for cracks in metal parts you’re often better off using eddy current. It can be slow and does need relative motion between part and sensor, but on the other hand, it’s a well-proven technology.

There was an informative eddy current application story published on the Quality Digest web site back on June 8th, “Crack Testing of Hypodermic Needles.” Could machine vision have done the job? Perhaps, but I’m pretty sure the error rate would have been much higher than with eddy current.

Just be aware, we might love machine vision but that doesn’t make it right for every problem.

And finally, an apology to those who clicked the link I provided yesterday to "hottestdigitalcameras" blog. Apparently the blog has been suspended so the link no longer takes you to the article. Sorry about that. Perhaps I should concentrate on providing more original material and fewer links. What do you think?

What you need to know when choosing a smart camera

One of my original goals for this blog was to provide a series of primers for people who needed to put a machine vision system to work quickly. Over time I’ve tended just to offer up links to where other people have done the same – it saves me time – as I figure there’s no point reinventing the wheel.

So here’s the latest link: if you need to quickly get a grasp of what’s important in a smart camera, check out,. “Considerations when choosing a smart camera” on the ‘Hottest Digital Cameras’ blog. Just one word of caution: there’s nothing about when a smart camera is the best tool for the job, but you have questions on that I’d be glad to help.

Non-manufacturing machine vision opportunity

Way back when I was a child I recall my mother reversing over my brother’s bike, abandoned on the driveway behind her car. Apparently it was my fault because I didn’t tell her – so much for parental justice, eh?

Well if she’d had a backup camera in her car (unheard of in the 1960’s, of course,) the crumpled frame might have been avoided. More importantly, if backup cameras were mandated in every vehicle, it appears that a lot of injuries and fatalities could be avoided. (“Oz Report Suggests Backup Cameras Should be Mandatory” – WardsAuto.com, 6/3/2101.)

Now I realize this is straying from my mission of providing aid to users of machine vision, but I bring it to your attention because it suggests there’s an absolutely huge machine vision opportunity here. Should backup cameras be mandated I imagine it will be the people who make cellphone cameras who will get the business. However, this will just put an image on the dash-mounted screen. What’s needed is some intelligence to analyze the image and warn the driver where there is a potential problem.

That means someone needs to develop an image analysis application specific to this problem and market. (And please don’t tell me we can depend on the phone-using, coffee-drinking driver to actually check the screen, because we know that’s about as reliable as human inspection.)

It also occurs to me that this system needs to work in darkness as well as daylight, so perhaps there’s also a need to build in an IR illuminator.

Some 60 million cars and light trucks are manufactured every year, so this is quite an opportunity.

Point Grey Video from The Vision Show

Vision Systems Design have made available a number of videos they shot at The Vision Show last month. While they are clearly all intended to promote the wares of the companies who paid for them that doesn't mean they can just be dismissed as advertising. There's some good information in there too.

A case in point is the presentation given by Vlad Tucakov of Point Grey. He runs through a number of their new products before wrapping up with something I think is genuinely interesting: the combination of the Point Grey Bumblebee stereo camera with 3DLocate from Cognex.

Watch the video by clicking this link.

Omron machine vision application

I have to admit that Omron is not the first name that springs to mind when I’m asked to list manufacturers of smart cameras, but it would be a mistake to overlook them completely. As this blog posting, (“Omron Vision Solution Improves Maruchan’s Uptime by Matching Lids to “Instant Lunch” Noodle Cup Contents”) discusses, they have vision products for less challenging applications.

The story itself is well-written and informative, so if you’re interested in vision applications, take a few minutes to read it. Unfortunately there are no photos, which would have helped explain the actual implementation, but what can you expect from a blogger?

Attendance stats for The Vision Show

Despite my absence (or possibly, because of my absence,) Jeff Burnstein, President of the AIA, was “thrilled with the show…” held in Boston last month. According to the press release, “Nearly 1,900 people registered…” which apparently is double the number from 2009.

I’m glad to hear trade show traffic is up because that suggests there’ll be more actual purchasing following on behind, but I can’t help thinking that if 1,900 visitors constitutes a successful show, then the AIA have set the bar rather low.

If you figure it cost each exhibitor $10,000 for their pitch (and I’m probably being conservative, especially if you factor in hotels and meals,) then they paid $5.26 for each pair of eyeballs. Does that represent good value for money?

If you think it does, then I will gladly accept your advertising on my blog, at the bargain rate of $5.00 per page view.

Machine vision lens calculator updated

For some time now I’ve been touting the free “MachVis” software tool from Linos as the best lens calculator around, so it’s only fair that I let you know they’ve tweaked it a little.

Version 3.6.1 of their “machine vision calculation software” adds a couple of new Rodagon lenses and tidies the display of sensor dimensions as well as fixing a couple of bugs. Not big changes, but if you’re a regular user it would make sense to stay current.

One point to note: obviously Linos are using MachVis as a tool to help sell their – very good – lenses, but you can of course just take the output from the program and apply it to products from competitor lens companies. MachVis doesn’t marry you to Linos.

And to think that they don’t even buy advertising on this blog!

Machine vision in surveillance

In the main I restrict this blog to industrial applications of machine vision, but on several occasions I’ve mentioned that security is a “target-rich environment” for developers of vision technology, so let’s take a short detour in that direction.

Surveillance relies to a great extent on the use of “pan-tilt-zoom” (PTZ) cameras. This results in lots of overlapping images with different scale factors, which are difficult to stitch together into a single scene. Researchers Marcos Nieto and Luis Salgado have been working on this problem and in “Automatic video mosaicing for surveillance using vanishing points” (SPIE web site, March 17th, 2010,) they set out a rather interesting approach.

As the title of paper suggests, their trick is to use vanishing points – those lines that disappear into the distance, meeting at some far point – as a way of registering disparate images. And guess what – it seems to work! (Check the paper to see the images.)

As a huge fan of the cult 1971 movie, “Vanishing Point,: I can’t help but be attracted to this method, even though Nieto and Salgado seem not to using a Dodge Challenger in their work. Check out this imdb link if you don’t know what I’m rambling about.

Now that’s high resolution!

Cypress Semiconductor have announced a 25 megapixel CMOS sensor capable of 53 frames per second (“Cypress Introduces Industry’s Highest Throughput Pipelined Global Shutter CMOS Image Sensor.” EDACafé blog, June 3rd, 2010)

How long will it be before someone incorporates it into a machine vision camera? I have a feeling we’ll see a 25Mp camera advertised before year end. It won’t be cheap though. My camera price rule of thumb has long been $1,000 per megapixel, although that does seem to be shifting downwards at the moment. Even so, my guess is that a 25Mp camera will be priced at a whisker under $20,000.

That might not sound like a bargain, but it depends on your application. For traffic and security applications a sensor like this provides terrific ability to zoom in on fine details while simultaneously capturing a large field of view. (Just like the way, in cop shows and movies, the hero will ask the resident nerd to “zoom in – now enhance” to produce a beautifully clear image of a license plate.)

As for my industrial, and price sensitive, applications – I guess I’ll be sticking with sensors of 1600 x 1200 pixel format for now.

Would collimated light help?

Well it might, but only if you know what it is, so go read “Measure Precisely with Machine Vision” by Ben Dawson of Dalsa and published on the Vision & Sensors web site (November 30th, 2009. Yes it’s an oldie, but it’s a goodie too.)

Don’t have the time or energy to click on the link I provided? OK, well let me summarize. You know how a telecentric lens improves the performance of a machine vision gauging system? (If not, search my blog for “telecentric.”) Well a collimated light complements the telecentric lens. Collimated means that all the light rays are travelling in the same direction, so you don’t get stray reflections off surfaces that are not facing the camera.

A good example would be when you want to backlight a shiny metal cylinder – when light rays are not aligned it makes the edge less sharp. Collimated light avoids this problem.

You can achieve nearly the same thing by placing the backlight at a distance from the object you’re gauging – the distance should be 2x or 3x more than the diagonal of the backlight so it becomes more of a point source – but it still won’t be quite as good. For ultimate precision you will need a collimated backlight.

More on the machine vision business environment

“Vision(s) of the Future,” on the Inspect-online blog (posted May 31st, 2010) discusses the direction of the machine vision business over the next ten years. There are some interesting thoughts, most of which I agree with, but the part that really makes me smile is where they discuss growth forecasts for 2010.

The blog notes wryly that while the EMVA anticipates double digit growth this year, the AIA is predicting 3%. As I’ve observed before, the data would suggest double-digit growth, meaning that the AIA are being extremely cautious.

Does this matter?

Well yes, I think it does. If manufacturers and distributors buy-in to the AIA numbers they are going to find themselves out of inventory. That’s going to create additional business for better stocked companies dealing in products for which customers have the ability to switch to vendor. However, when customers are locked in to a single supply it’s just going to mean extended delivery quotes – something we’re seeing already.

The danger then is that a cautious forecast will become a self-fulfilling prophecy (as well as an opportunity to ramp up prices.) Hopefully the AIA will look again at its forecast. If not, I think we could talk ourselves out of recovery.

No more machine vision “evals”?

I think it was Jon Chouinard, then working for NER, who coined the phrase “poke and hope” in describing how the typical machine vision feasibility consists of much trial and error and very little science.

While I agree with Jon that we should use science and engineering to arrive at solutions to our machine vision challenges, sometimes it’s still necessary to try stuff out and see what works.

This is where it’s so useful to be able to borrow hardware for evaluation. I don’t care how well equipped your bench or lab is, there will be times when you haven’t got the tool that seems most likely to solve your problem. So you do what I’ve always done, which is to call up the manufacturer or their distributor and ask nicely if you can borrow said tool.

This worked well back in the ‘olden days’ (2009) but it isn’t working so well now. I’m starting to find that some companies – no, I’m not naming names – are refusing to provide evaluation units. Instead they say, “If you want to try our camera you’ll have to buy it.”

I understand that they have a business to run, but since higher-end cameras will run several thousand dollars, and I’m still under pressure to manage my costs, that’s a non-starter for me. There are other, more helpful, sources of hardware, but I’m not happy with the way my choice is being limited.

Fortunately this unhelpfulness is restricted to just a few companies at the moment, but since the beginning of the year I have noticed other companies getting a lot more formal about how they manage their evaluation/loaner program. I can’t blame them for that: we are talking about some high dollar kit, and when it’s been used once it only has secondhand value. But telling me to buy rather than borrow doesn’t help me and it sure as heck isn’t going to help them sell more cameras.

Don’t overlook Matrox

Cognex have done a terrific job of making their InSight smart camera range the tool of choice for any manufacturing engineer needing to implement machine vision. The yellow housings are instantly recognizable and engineers know the software will work. Back in the ‘70’s ‘they’ used to say that no one ever got fired for buying IBM. Well the same goes for Cognex in the machine vision world: it’s the safe choice so why shop around?

Well before handing over your credit card to the folks in Natick (or one of their distributors,) I suggest you do a little “due diligence” and take a look at the Matrox Iris GT smart camera. OK, so a camera is a camera, but the software merits a closer look. Matrox have been developing their “Design Assistant” package to the point where it’s a worthy alternative to InSight Explorer. The latest release, 2.3, (due out any day) improves upon their part fixturing capabilities and adds a bead tool, which I think is quite useful. In addition, and I realize this is just personal choice, I prefer the flowchart approach to programming/development taken by Matrox in Design Assistant to the InSight spreadsheet. Your experience may of course be different.

That’s all for now. I just wanted to say, don’t automatically plump for Cognex. You owe it to yourself (and your employer,) to shop around.