Los Alamos GENIE mimics evolution to get at complex features in digital images
Named GENetic Imagery Exploitation, GENIE's ability to evolve superior algorithms allows it to find the features of interest in nearly any set of images. GENIE can be used to map damage caused by natural disasters such as wildland fires, hurricanes, floods, earthquakes and volcanoes or man-made ones such as terrorist attacks. It also can monitor environmental changes or crop health; track population growth or detect signs of disease in medical images.
The system is a 2002 R&D 100 Award winner.
Sorting through volumes of digital images
Facing the system's developers was a complex issue in need of a solution – the staggering volume and variety of digital images available to scientists. These images come from any number of different sources. Consider a few that come from space -- Hubble Space Telescope images of the most distant objects in the universe, aircraft and satellite images of the entire surface of Earth taken in different spectral bands, images taken by NASA's lunar and Martian orbiters and rovers, and spacecraft images of other planets in the solar system as well as asteroids. Closer to home are medical images, such as conventional X-rays, biopsy micrographs and CAT, MRI, PET and ultrasound images. Then, there are aerial views of wildfires, complex ecosystems and geological formations among others.
GENIE assembles an initial set of low-level image-processing algorithms (for example, edge detectors, texture measures and spectral operators) and then tests each algorithm's ability to find the feature of interest.
"Less fit" algorithms are discarded while the "more fit" are combined to produce superior ones. After several generations of survival of the fittest, the resulting algorithm is highly optimized. Although features and imagery constantly change, GENIE's ability to evolve superior algorithms allows it to find the features of interest in nearly any set of images, according to the development team.
Understanding complex tasks
GENIE has been used successfully to find and map damage caused by wildfires, pollution and terrorist attacks. Future applications could include other disaster scenarios such as hurricanes, floods, overgrazing and habitat loss, insect infestations and volcanoes.
But GENIE is not limited to mapping ecosystems and monitoring important environmental change, its developers say. Presented with other, very different types of imagery, it is showing an ability to detect signs of disease in medical images and to map craters and other surface features on distant planets. In the future, it might assist baggage screeners at airport security checkpoints and be used to detect defects in products made on assembly lines.
Its own results can be reused to help GENIE build up its own "understanding" of complex tasks. For example, after GENIE learns to find water, it then can learn to find beaches. GENIE learns to ignore unimportant image-to-image variations such as atmospheric haze or variations in overall illumination.
Principal developers at Los Alamos are Jeffrey Bloch, Steven Brumby, Nancy David, Mark Galassi, Neal Havey, Simon Perkins, James Theiler, A. Cody Young, Diana Esch-Mosher, Reid Porter and John Szymanski
Los Alamos National Laboratory is operated by the University of California for the National Nuclear Security Administration (NNSA) of the U.S. Department of Energy and works in partnership with NNSA's Sandia and Lawrence Livermore national laboratories to support NNSA in its mission.
Los Alamos enhances global security by ensuring safety and confidence in the U.S. nuclear stockpile, developing technologies to reduce threats from weapons of mass destruction and improving the environmental and nuclear materials legacy of the cold war. Los Alamos' capabilities assist the nation in addressing energy, environment, infrastructure and biological security problems.
For more Los Alamos news releases, visit http://www.lanl.gov/external/news/releases.