Tuesday, February 10, 2009

Should Cognex be worried?

Are “smart cameras” are overpriced? If you consider that they are little more than an image sensor, a DSP chip and some software, then perhaps $2,000 and up is expensive. On the other hand, the volumes are low and the manufacturers need to provide a high level of after-sales support, so maybe $2,000 is a bargain.

Apparently Reto Baettig, head of Vision at “Supercomputing Systems” gave this question some thought before coming up with the “
LeanXcam.” This won the “2008 Vision Award” at the big Stuttgart show last year, so it seems worth closer examination.

According to “
Test & Measurement World,” this is an open-source smart camera with a cost of under €200. (Cost to build or price to buy? The article is unclear.)

“But what,” I hear you asking, “Is an open-source smart camera? And why would anyone want to offer such a thing?”

These are good questions, so let me summarize my understanding of the product, and then share some thoughts as to “Why.”

Image acquisition is by means of a 752 x 480 pixel CMOS sensor that runs at 60fps. Images are processed by a Blackfin DSP and communication is by Ethernet or camera I/O. The camera’s operating system is
μClinux and the development environment utilizes the OSCar (Open-Source Camera) Software Framework. Apparently the intention is that a “community” will develop a library of programs freely available for all users.

Now I can certainly imagine a small army of university researchers putting the LeanXcam to work in various robot guidance and biologic imaging applications, but is anyone in industry going to use it? The problem I see is that industrial users of machine vision want image processing and analysis tools that are easy to apply, preferably with some kind of wizard to guide application development. The OSCar Framework may eventually offer a wide library of tools, but will they all be simple to use, documented, and most importantly, reliable? And will there be support for when things go wrong?

And while I’m being Mister Negative, what do the LeanXcam and OSCar offer that I can’t get from Roborealm? Yes, I understand it’s a single, low-cost package, and that might offer some advantages, but do I want to take the chance on a new product when something similar already exists?

This just leaves us with one question: what’s the real motivation for the LeanXcam? I suggest that perhaps this is a way for “Supercomputing Systems” to garner world-wide exposure for the considerable expertise they clearly posses, and thereby sell their services to those looking to develop their own proprietary “smart cameras.”

So to the question posed in the title of this post, no, I don’t think Cognex need loose any sleep over the LeanXcam. It’s an interesting idea but I don’t expect to see swarms of the little boxes in factories any time soon.

2 comments:

Andy Wilson said...

Dear Sir:

Re: Swarms of these cameras are already appearing. Ferag AG (Hinwil/Zürich, Switzerland; www.ferag.ch), a manufacturer of conveying, binding, inserting, and stitching machines for the newspaper and magazine post-processing industry, for example, is using an industrial version of the camera to monitor the correct folding of magazine forms before they are finally bound. For more information, go to: http://www.vision-systems.com/display_article/349698/19/none/none/Techt/IMAGE-PROCESSING:-Open-source-system-wins-VISION-Award

Anonymous said...

Sure Cognex should be worried. The factory automation segment is heavily dependent on distributors for marketing and post-sales support and start-up. This channel is being squeezed by other Cognex products that they are not allowed to sell. Their margins are falling. Cognex overpaid for DVT's distribution channel and will never get it to work.

Cognex is very vulnerable to a technology that could be viewed as attractive to the mavericks of tomorrow's online inspection techniques.

The one saving-grace for Cognex is that it is sitting on a lot of cash. It needs to do more than strike strategic alliances with companies like Mitsubishi. Cognex needs to position itself (acquire) technologies that serve outside of semis long-term and finally define its go to market strategy for the factory automation sector.

If not, look for Rockwell or Siemens to purchase them before Dr. Bob's shares are worthless in the next 24 months. (Bob would be a complete full if he weren't at least entertaining this option).