11/12#03
VISUAL ROBOTICS
Yale University
| 0941 |
| |
(0941)
DID YOU KNOW THAT ROBOTS ARE NOW ALMOST CAPABLE OF SEEING?
THAT'S RIGHT,
COMPUTER WHIZZES ON THE YALE CAMPUS ARE LINKING COMPUTERS AND ROBOTS WITH
CAMERAS THAT ACT LIKE EYES.
| 0953 |
| |
(0953)
DR. GREGORY HAGER OF YALE UNIVERSITY IS
ACTUALLY PAID TO SIT IN FRONT OF A
COMPUTER SCREEN AND MAKE SILLY FACES. BUT WHAT'S GOING ON HERE IS MORE
THAN WHAT MEETS THE EYE. THERE'S A CAMERA
NEAR HIS COMPUTER THAT'S OBSERVING HIS EVERY MOTION.
| 1009 |
| |
(1009)
The computer is seeing only in a very restricted sense.
We've told it,
for example, if it's going to track a face, than what it wants to do is
find an eye for example in one image. And in the next image, find that
same eye again. It doesn't understand anything about eyes. It merely
understand that there is an an object with a particular appearance and in
subsequent images, it's trying to find the same object with roughly the
same appearance.
(1036)
THE IDEA OF LINKING COMPUTERS WITH CAMERAS IS NOT NEW.
FOR INSTANCE, BASIC TELECONFERENCING CAN ALREADY LINK PEOPLE TOGETHER INTO
A LONG DISTANCE MEETING AS LONG AS THE PARTICIPANTS STAY SEATED DIRECTLY IN
FRONT OF THE COMPUTER'S CAMERA.
| 1056 |
| |
(1056)
BUT WITH HAGER'S VISUAL TRACKING TECHNIQUE, THE CAMERA ACTUALLY HAS A
ROVING EYE AND CAN FOLLOW A TARGET THAT HAS BEEN MARKED, LIKE THE EYES AND
MOUTH ON THE CLOWN.
(1105)
THIS MEANS THAT A LECTURER CAN GET UP,
MOVE AROUND, MAKE A DEMONSTRATION
AND THE CAMERA WOULD CONTINUE TO FOLLOW THE SUBJECT AND PROJECT THE IMAGE.
ACTORS COULD FILM THEMSELVES
AND THE LIST OF POSSIBILITIES GOES ON...
| 1118 |
| |
(1118)
BEYOND FACES, COMPUTERS WITH CAMERAS CAN
ALSO TRACK OTHER THINGS AND ALLOW
FOR ALL SORTS OF MOVEMENT.
| 1125 |
| |
(1125)
This is our surf ball and that camera over there is watching this ball
which has special markings on the other side and using the information
contained in those markings, it determines that now I'm moving the ball.
| 1137 |
| |
(1137)
IMAGINE A COMPUTER MOUSE THAT MOVES IN THREE DIMENSIONS.
THAT'S EXACTLY THE IDEA BEHIND THIS SURF BALL--UNLIKE THE
STANDARD MOUSE ON YOUR COMPUTER WHICH ONLY MOVES UP AND DOWN, LEFT AND
RIGHT, THE SURF BALL ALLOWS YOU TO MOVE IN THREE DEGREES OF FREEDOM IN
ROTATION.
(1153)
NOW ATTACH THIS FLEXIBLE VISUAL SENSOR TO A ROBOT
AND YOU'VE GOT A ROBOT
WITH A SET OF EYES THAT HAS DEPTH PERCEPTION, MUCH LIKE HUMAN BEINGS.
(1203)
The computer interprets the ball's motion and
sends commands to the robot
that mimic my motions. For example, if I push forward, the robot moves
forward. If I pull back, it moves backwards, and similarly side to side
and up and down.
(1218)
THIS SAME TECHNOLOGY COULD BE USED TO
CONTROL A VIRTUAL REALITY ENVIRONMENT
OR ANY TYPE OF ENVIRONMENT THAT YOU NEED TO MOVE IN ALL DIRECTIONS.
| 1226 |
| |
(1226)
The next generation of computers are going to be fundamentally
three-dimensional. Graphics will go from the two-dimensional doman to full
3-D perception. And when you do so, there are undoubtedly any number of
applications.
(1241)
For example, imagine a drawing program in three dimensions.
And you have to now manipulate and rotate the structures in that space. It
would be very natural to have a device like this that lets you, for all
practical purposes, latch onto the object you're trying to manipulate, and
then have the sense of manipulating it directly.
(1303)
RIGHT NOW ROBOTS WITHOUT EYES LIKE THESE
IN A CAR FACTORY ARE LIMITED TO
PLAYING BACK A SPECIFIC MOTION OVER AND OVER. PUT THEM IN AN ENVIRONMENT
WHERE THEY NEED TO FIND OBJECTS AND THEY'RE LOST.
(1315)
BUT THE POSSIBILITIES FOR SEEING-EYE ROBOTS BECOME EXPLOSIVE WHEN YOU
CONSIDER THAT THEY--INSTEAD OF HUMANS--
COULD BE OUT HERE IN BOSNIA DISMANTALING LAND MINES OR IN A ONLY ROBOTS
VARIETY OF DANGEROUS PLACES. WITH THAT KIND OF FLEXIBILITY, THE ONLY RISK
TO LIFE OR LIMB.
Copyright (C) 1996 Japan High-Tech Satellite Network Co., Ltd., NEC Corporation
All Rights Reserved. Reproduction of the information hereby
is strictly prohibited without written consent.
