Mosaic of the near side of the moon as taken by the Clementine star trackers. The images were taken on March 15, 1994
For NASA researchers, pixels are much more – they are precious data
that help us understand where we came from, where we've been, and where
we're going.
At NASA's Ames Research Center, Moffett Field, Calif., computer
scientists have made a giant leap forward to pull as much information
from imperfect static images as possible. With their advancement in
image processing algorithms, the legacy data from the Apollo Metric
Camera onboard Apollo 15, 16 and 17 can be transformed into an
informative and immersive 3D mosaic map of a large and scientifically
interesting part of the moon.
The "Apollo Zone" Digital Image Mosaic (DIM) and Digital Terrain
Model (DTM) maps cover about 18 percent of the lunar surface at a
resolution of 98 feet (30 meters) per pixel. The maps are the result of
three years of work by the Intelligent Robotics Group (IRG) at NASA
Ames, and are available to view through the NASA Lunar Mapping and
Modeling Portal (LMMP) and Google Moon feature in Google Earth.
"The main challenge of the Apollo Zone project was that we had very
old data – scans, not captured in digital format," said Ara Nefian, a
senior scientist with the IRG and Carnegie Mellon University-Silicon
Valley. "They were taken with the technology we had over 40 years ago
with imprecise camera positions, orientations and exposure time by
today’s standards."
The researchers overcame the challenge by developing new computer
vision algorithms to automatically generate the 2D and 3D maps.
Algorithms are sets of computer code that create a procedure for how to
handle certain set processes. For example, part of the 2D imaging
algorithms align many images taken from various positions with various
exposure times into one seamless image mosaic. In the mosaic, areas in
shadows, which show up as patches of dark or black pixels are
automatically replaced by lighter gray pixels. These show more well-lit
detail from other images of the same area to create a more detailed map.
Left: A normal one-camera image of the lunar surface. Right: A composite
Apollo Zone image showing the best details from multiple photographs.
"The key innovation that we made was to create a fully automatic image
mosaicking and terrain modeling software system for orbital imagery,"
said Terry Fong, director of IRG. "We have since released this software
in several open-source libraries including Ames Stereo Pipeline, Neo-Geography Toolkit and NASA Vision Workbench."
Lunar imagery of varying coverage and resolution has been released
for general use for some time. In 2009, the IRG helped Google develop
"Moon in Google Earth", an interactive, 3D atlas of the moon. With "Moon
in Google Earth", users can explore a virtual moonscape, including
imagery captured by the Apollo, Clementine and Lunar Orbiter missions.
The Apollo Zone project uses imagery recently scanned at NASA's
Johnson Space Center in Houston, Texas, by a team from Arizona State
University. The source images themselves are large – 20,000 pixels by
20,000 pixels, and the IRG aligned and processed more than 4,000 of
them. To process the maps, they used Ames' Pleiades supercomputer.
The color on this map represents the terrain elevation in the Apollo Zone mapped area.
The initial goal of the project was to build large-scale image
mosaics and terrain maps to support future lunar exploration. However,
the project's progress will have long-lasting technological impacts on
many targets of future exploration. "The algorithms are very complex, so they don't yet necessarily apply to
things like real time robotics, but they are extremely precise and
accurate," said Nefian. "It's a robust technological solution to deal
with insufficient data, and qualities like this make it superb for
future exploration, such as a reconnaissance or mapping mission to a
Near Earth Object."
Near Earth Objects, or "NEOs" are comets and asteroids that have been
attracted by the gravity of nearby planets into orbits in Earth's
neighborhood. NEOs are often small and irregular, which makes their
paths hard to predict. With these algorithms, even imperfect imagery of a
NEO could be transformed into detailed 3D maps to help researchers
better understand the shape of it, and how it might travel while in our
neighborhood.
In the future, the team plans to expand the use of their algorithms
to include imagery taken at angles, rather than just straight down at
the surface. A technique called photoclinometry – or "shape from
shading" – allows 3D terrain to be reconstructed from a single 2D image
by comparing how surfaces sloping toward the sun appear brighter than
areas that slope away from it. Also, the team will study imagery not
just as pictures, but as physical models that give information about all
the factors affect how the final image is depicted.
"As NASA continues to build technologies that will enable future
robotic and human exploration, our researchers are looking for new and
clever ways to get more out of the data we capture," said Victoria
Friedensen, Joint Robotic Precursor Activities manager of the Human
Exploration Operations Mission Directorate at NASA Headquarters. "This
technology is going to have great benefit for us as we take the next
steps."
From physorg
0 comments:
Post a Comment