Tech Thursday: Visionary Machines

Permalink 03/03/05  

Back in the 1960s HAL 9000's glowing red "eye" served as one of the first realistic ideas of machine vision (as opposed to Gort from "The Day The Earth Stood Still", and others). But the reality was that machine vision was barely more than simple light beam interrupt sensors -- the kind that you can now find in products as simple as children's toys. Now, almost 40 years later, the technologies once only dreamed of are allowing real products to see for the first time. Tech Thursday looks at the new opportunities opened up by these artificial eyes.


How many times have you wished that a product, like a garbage can, could just see you coming and open up, or turn on, or maybe even just cook you dinner? Before computer vision systems began cheapening up in the last few years, these were pipe dreams, but now, we are approaching a time where machines will be able to "see" a wide variety of information, and make decisions based on it. But, even the human eye is broken up into different areas for different purposes; peripheral vision is designed to detect motion and shadow information, while your central vision fills in the detail information in the brain. Machine vision is no different, and each different system has strengths at a special facet of vision. Take a look at your options:

Eye Tracking
Obviously, having machines that know where users are looking gives a lot of information about the person's train of thought. Most systems that do this are used for ethnographic research into where people look in order to perform different tasks, like browsing web pages. In most cases, like this Polyhemus system, one or two infrared-illuminated digital cameras take video of the eyes, and position of looking is calculated by a computer. You can see a demonstration of the output of this kind of vision over a webpage for usage-tracking here. Some companies have taken it a step further, and integrated the cameras into monitors for eye-based mouse control and clicking.




Body Part Tracking and Gesture Recognition
Not content to use just the eyes, some researchers are experimenting with other more easily tracked body parts, like the nose, which can't be covered up as easily by a turned head. Gesture recognition also is becoming a big deal, since so much of our communication is done through these subtle hand and body movements. Gestures allow gross navigation of a computer, freeing up voice-type commands for the fine details like navigating file structures and submitting searches. Asimo is even being designed to incorporate gesture-based commands.




Facial Recognition
separate from watching movement of bodies, some computers, like those searching huge networks of surveillance cameras for certain people, can benefit from being able to recognize faces. Facekey is one of the many companies which has commercialized a facial recognition system to run on a PC. In addition, Omron semiconductors makes a chip which brings facial recognition and biometric security to cellphones. Even though surveillance network facial recognition isn't quite up to par yet, it's definitely on its way, and for better or worse, we'll get there someday. More facial recognition info is available from RAND. Also, see Carnegie Mellon's Face Group at the Robotics Institute for lots more research.




Movement or Edge Tracking
This is much more basic than any of the above methods, and is the most used method in current games and toys like the Ion game system, and EyeToy gaming system for Playstation 2. Logitech also uses it in it's latest webcam, which can pan to follow you around a room. This method only lets the machine know that something is moving, or that the edge of your hand or arm is there, but not whether it is actually your hand, your leg, or whatever, so that must be accounted for in the design.




Navigation Vision
Vision systems that control vehicles or robots themselves are probably the ultimate goal of the whole machine vision movement, and they are getting more and more incredible. Despite the lackluster performances of teams at the Darpa Grand Challenge last year, the vision systems used were unbelievable. Most of these systems use combination of a lot of very complicated algorithms, and incorporate compass and GPS information as well. For more on Autonomous Vehicles, check out the American Association for Artificial Intelligence.



Conclusion:
With computing getting so cheap and omnipresent (For crying out loud, it's facial recognition in cellphones!) there are bound to be some amazing new products and experiences built around devices that see in the next few years. Maybe you'll be the one to make them. It's definitely worth keeping in mind.
PermalinkPermalink

Previous post: Humansys: Hot Mammas (and Poppas)

Next post: Idealism VS Implementation

 

 

Copyright 2004-2006 Dominic Muren and IDFuel Team




Advertise on IDFuel

Technorati Profile

Search




Categories

Archives

XML Feeds
RSS 1.0
RSS 2.0
Atom
 What is this?