The Intelligent Robotics Group has created a “Crater Detection” challenge for the NASA Tournament Lab. This challenge focuses on detecting impact craters in orbital images and will run from June 15-29, 2011. Detecting these features is of keen interest to mission planners and scientists at NASA. Large-scale automatic and robust crater detection algorithms can help solve one of the most challenging and important data mining problems in space exploration.
As the most common topographical feature, craters provide important information on planet formation and geology. Craters also inform the selection of landing sites, provide valuable data for path planning and rover navigation, and help scientists align a variety of disparate data sets (radar, laser altimetry, etc.) to each other.
The data for this challenge consists of orbital images captured by the Apollo 15-17 and Lunar Reconnaissance Orbiter missions. Craters were manually identified by NASA planetary scientists and by volunteers from around the world as part of the Moon Zoo project. The manually labeled craters provide both training data and a way to measure the performance of challenge submissions.
Moon Zoo is a Zooniverse project. Zooniverse has over 430,000 registered users helping on a wide variety of Citizen Science projects. The data provided by Moon Zoo for this challenge represents a small sample of a much larger crater catalog currently being compiled. The first full crater catalog from Moon Zoo should be available in late 2011.
Overall, this challenge will help NASA to better, and more effectively, process the ever-increasing amount of orbital imagery of the Moon, Mars, and other planetary bodies. This will enable us to better explore and understand our solar system and beyond.
BACKGROUND: To encourage and foster open innovation, NASA established the NASA Tournament Lab through a contract with Harvard University and TopCoder. The NTL enables software developers to compete with each other to create innovative computer code for NASA systems. The NTL provides an online virtual facility for NASA researchers with a computational — or complex data processing — challenge to “order” a solution, just like they would order laboratory tests or supplies. This approach, often called “crowd sourcing” or “broadcast search,” can improve the quality and speed of software development by searching for a solution to a problem through multiple, parallel paths. Instead of relying on one individual or team, NASA can access many independent ideas, increasing the chances of finding a successful solution.
NASA PROGRAM FUNDING: NASA ESMD Lunar Mapping and Modeling Project
Contact: Terry Fong