EE Times: Like what?
Bier: For example,
a camera can tag that this is a picture of Barbara and Fred, and then find
the other pictures you’ve taken recently that include those same
EE Times: Hmm. That may be useful for Facebook posting… What about embedded vision apps for systems other than handsets and cameras?
Somewhat counter-intuitively, vision-based automotive safety
applications are also very power-constrained. Of course, there’s a big
battery available in the car, but the issue is getting the heat out.
You might have Advanced Driver Assistance Systems (ADAS) completely
contained in a rear-view mirror assembly, for example, and that sucker
gets hot when you’re car’s been sitting all day in the parking lot in
Phoenix in August. But when you turn the key, it’s got to work 100 percent of
This means you need a very low-power system that
dissipates very little heat. And ADAS is coming on big. For example, the 2013
Honda Accord is the first non-luxury car to have vision-based safety
features as a factory option.
EE Times: The power issue of ADAS is something I had not thought about before.
Bier: OK, one more and then I’ll stop. You know that LiveScribe “smart pen” you have? [Full disclosure: Bier gave me this LiveScribe pen a year ago.]
That thing runs all day on a tiny battery, and it’s doing computer
vision every time you write with it--that’s how it tracks pen strokes
and identifies which page you’re on.
EE Times: In
other words, when you have to keep capturing images and
processing/tracking at the same time, you do need a low-power solution.
I am shifting a gear here. Do you have any market data that shows a
growth rate of mobile devices that are equipped with image/video
Bier: Let’s start with smartphones
and tablets. As you know, these are now very high-volume products, and
they’re incorporating more and more embedded vision functionality. [Bier
pointed out that in the third quarter alone last year, 169.2 million
units of smartphones were sold worldwide, according to Gartner. During
the same period, 27.8 million units of tablets were shipped worldwide,
according to IDC.] Dual and quad-core CPUs (Cortex-A9-class and
above) are often designed into such tablets, making them beefy embedded
EE Times: What about the embedded vision growth rate of automotive?
Yes, ADAS. This is currently a small market, but it’s going to be
huge in my opinion, due to three factors: 1) Something like 100 million
cars and light trucks are manufactured each year; 2) Car accidents cause
an incredible amount of loss of life, injury and property damage
(they’re the No. 1 cause of accidental childhood death in the U.S.). 3)
ADAS systems can increasingly be made affordable, thanks to improvements
in processors, sensors and other technologies.
Now, it’s still
early adopter days now and manufacturers are charging a premium for
these system, just like they did for airbags, anti-lock brakes, and
stability control in the early days of those technologies.
Today, virtually every car sold in the U.S. has those features. According to IMS Research, vision-based ADAS systems will become a billion-dollar-a-year business in the next two years or so.