Light show makes 3D camera
By
Kimberly Patch,
Technology Research News
True three-dimensional movies, where projected
characters look like they are occupying the same room as the audience,
require a couple of new technologies: a projection system that can put
the many dots of color that make up the three-dimensional image in the
right places at the right time, and a camera that can capture the action
in the first place.
Researchers from the University of Kentucky have come up with
a relatively low-cost method to measure depth using a single camera. The
scheme involves shining a light pattern onto an object, and gaining depth
information from the way the object distorts the pattern.
The researchers' depth camera prototype is made from off-the-shelf
parts worth about $4,000, Chun Guan, a researcher at the University of
Kentucky. "A system well below $1,000 is certainly possible," given mass
production, he said.
The camera could lower the cost of computer vision systems that
enable computers to locate people and sense gestures, and could eventually
be used to capture the depth information needed for three-dimensional
videos.
Existing computer vision systems that identify a person's location
and read gestures use several cameras positioned at different angles to
triangulate depth information. The researchers' method requires one camera,
and does not use a lot of compute power, said Guan. "Structured light
imaging has several benefits [including] lower computational cost to extract
the depth video from the raw recorded data," said Guan.
The light patterns are comparable to sunlight coming through Venetian
blinds and striping objects in a room, said Guan. "When viewed from an
angle, the stripes are crooked due to the objects, and this distortion
can be used to calculate the three-dimensional shapes" of the objects,
he said.
The problem with using just one set of stripes is that when an
object has an edge, the computer will lose track of which stripe is which.
The key to the scheme is that the researchers use several different patterns
of light and triangulate to gain full three-dimensional information. "Multiple
patterns are... required to achieve non-ambiguity and good depth resolution,"
said Guan.
The researchers' device projects a composite image, separates
out the individual patterns, and triangulates to determine the depth information
just as if the patterns were projected and captured separately, said Guan.
The setup captures depth information quickly -- as fast as the camera
image can be digitized, he said.
Previous structured light depth imagers have used multiple patterns,
but had to process them one at a time. The researchers got around this
problem by making each pattern sweep sideways at a different speed. The
speeds are analogous to radio station channels, said Guan. "Each pattern
has its own carrier frequency, or channel," he said. The camera uses the
frequencies to separate and analyze the patterns simultaneously.
There's a long list of possible applications, according to Guan.
"We're going to investigate human-computer interaction and automated surveillance
applications," he said. The camera could also be used in manufacturing
to inspect objects from afar.
The researchers tested the light pattern method in a setup that
allowed people to change viewpoints in a virtual reality environment by
moving their hands backward and forward. They also used the scheme as
part of a computer interface that allows people to issue commands by making
the motions of pressing buttons.
The researchers are working on implementing the depth camera prototype
into a head and hand tracker that will allow persons with disabilities
to interact with computers, said Guan. "In this case, it is a primary
goal for work to keep the price tag of the device under the $4,000 threshold"
including the PC involved, he said.
Eventually the technology could find its way into consumer video
cameras, said Guan. In the movie Minority Report, Tom Cruise's character
watches a home video of his son on a holographic display where a full,
three-dimensional reconstruction of the child walks out of the background.
"We hope to build a camera that would be the recording device for that
scene," he said.
The researchers are working on combining the depth capture information
with regular video to make a true three-dimensional camera, said Guan.
The goal is to "move the structured-light source and the camera into the
near infrared range, and then to couple a standard [color video] camera
such that the resulting signal will be an RGB-plus-depth video signal,"
he said.
Such a camera could also be used instead of the many cameras required
today to obtain three-dimensional special effects like those seen in the
movie Matrix, said Guan. "Instead of using hundreds of single image cameras
to obtain 3D special effects... only three or four three-dimensional video
cameras would be necessary," he said. And "rather than a limited special
effect, a full surround 3D video format could be made that could be viewed
with [the] orientation controlled by the viewer," he said.
What the researchers have accomplished so far constitutes a good
approach to capture a rough depth map of a scene, said A. Ardeshir Goshtasby,
a professor of science and engineering at Wright State University. "This
is useful in robot navigation where [an] approximate depth map is sufficient,"
he said.
A practical three-dimensional video camera is possible within
two years, said Guan.
Guan's research colleagues were Lawrence G. Hassebrook and Daniel
Lau. The work appeared in the March 10, 2003 issue of Optics Express.
The research was funded by NASA.
Timeline: < 2 years
Funding: Government
TRN Categories: Computer Vision and Image Processing; Human-Computer
Interaction
Story Type: News
Related Elements: Technical paper, "Composite Structured
Light Pattern for Three-dimensional Video," Optics Express, March 10,
2003.
Advertisements:
|
May 5/7, 2003
Page
One
Screen arcs widen view
Light show makes 3D
camera
Net scan finds like-minded
users
Sound forms virtual
test tubes
News briefs:
Nanotube shines
telecom light
Touchy-feely goes remote
Light mix makes
strong metal
Metal expands electrically
Researchers
fill virus with metal
Gold connectors stretch
News:
Research News Roundup
Research Watch blog
Features:
View from the High Ground Q&A
How It Works
RSS Feeds:
News | Blog
| Books
Ad links:
Buy an ad link
Advertisements:
|
|
|
|