Feature Detection Systems Enhance Satellite Imagery
- Created: Sunday, 01 November 2009
Originating Technology/NASA Contribution
In 1963, during the ninth orbit of the Faith 7 capsule, astronaut Gordon Cooper skipped his nap and took some photos of the Earth below using a Hasselblad camera. The sole flier on the Mercury-Atlas 9 mission, Cooper took 24 photos—never-before-seen images including the Tibetan plateau, the crinkled heights of the Himalayas, and the jagged coast of Burma. From his lofty perch over 100 miles above the Earth, Cooper noted villages, roads, rivers, and even, on occasion, individual houses.
In 1965, encouraged by the effectiveness of NASA’s orbital photography experiments during the Mercury and subsequent Gemini manned space flight missions, U.S. Geological Survey (USGS) director William Pecora put forward a plan for a remote sensing satellite program that would collect information about the planet never before attainable. By 1972, NASA had built and launched Landsat 1, the first in a series of Landsat sensors that have combined to provide the longest continuous collection of space-based Earth imagery. The archived Landsat data—37 years worth and counting—has provided a vast library of information allowing not only the extensive mapping of Earth’s surface but also the study of its environmental changes, from receding glaciers and tropical deforestation to urban growth and crop harvests. Developed and launched by NASA with data collection operated at various times by the Agency, the National Oceanic and Atmospheric Administration (NOAA), Earth Observation Satellite Company (EOSAT, a private sector partnership that became Space Imaging Corporation in 1996), and USGS, Landsat sensors have recorded flooding from Hurricane Katrina, the building boom in Dubai, and the extinction of the Aral Sea, offering scientists invaluable insights into the natural and manmade changes that shape the world.
Of the seven Landsat sensors launched since 1972, Landsat 5 and Landsat 7 are still operational. Though both are in use well beyond their intended lifespans, the mid-resolution satellites, which provide the benefit of images detailed enough to reveal large features like highways while still broad enough for global coverage, continue to scan the entirety of the Earth’s surface. In 2012, NASA plans to launch the Landsat Data Continuity Mission (LDCM), or Landsat 8, to extend the Landsat program’s contributions to cartography, water management, natural disaster relief planning, and more.
In 2002, Geospatial Data Analysis Corporation (GDA), of State College, Pennsylvania, received a Phase I Small Business Innovation Research (SBIR) contract with Stennis Space Center. (The company also engaged in a follow-on Phase II SBIR with Stennis.) The NASA center was seeking a new method for detecting clouds for future use with Landsat 8. Cloud contamination is a common problem with satellite imagery. For previous Landsat missions, NASA has used its Automated Cloud Cover Assessment (ACCA) algorithm to estimate cloud contamination in Landsat images. ACCA heavily relies on thermal data from the Landsat satellites—thermal signals being the easiest method for cloud detection. Thermal sensors, however, are expensive additions to satellites, and Landsat 8 will not feature any in its array. GDA took on the challenge of developing a new system that could accurately detect clouds without the benefit of thermal data.
“We proposed that, in order to identify clouds automatically and without thermal data, you have to move beyond the spectral signal in the imagery and into spatial feature and pattern recognition,” says Dr. Stephanie Hulina, GDA president and senior scientist. GDA employed its SBIR funding to arrive at the embodiment of this approach: the Cloud and Cloud Shadow Assessment (CASA) software.