Machine Learning Meets Geospatial Big Data
May 31, 2016 By digitalglobe
James Crawford, Founder and CEO of Orbital Insight, has been interested in space for a long time. When he worked as robotics and artificial intelligence expert for NASA, Crawford pioneered AI support for spacecraft and observation satellites. Now, he has turned his attention to planet Earth, using machine learning to extract intelligence from Geospatial Big Data (GBD).
Crawford often describes Orbital Insight as a macroscope: a tool that can identify individual objects on the Earth’s surface, but then be able to scale that up to cover vast geographical areas. Using cloud computing and graphical processing units, Orbital Insight can see the forest and the trees. Deep learning and AI automate the processes to create scalable insights from large amounts of data.
Think of Orbital Insight as Google Books for satellite imagery. Google Books takes an image of each page of twenty million books and processes them so that, when a user searches “to be or not to be,” Shakespeare’s Hamlet pops up. At Orbital Insight, Crawford’s team takes millions of satellite images and processes them to answer questions like “What will the U.S. corn yield be this year?” and “Which Chinese cities are growing fastest?” To answer these questions, Orbital Insight needs a pipeline that lets them access and analyze geospatial big data. Orbital Insight’s choice: the cloud-based GBDX platform.
Usually, managing large amounts of data requires large amounts of storage. Not with GBDX. Orbital Insight has on-demand cloud-based access to DigitalGlobe’s archive of more than 70 petabytes of imagery, which grows by over 3,000,000 square kilometers of new imagery every day. Users can pull every available image from the library—both current and historical. Orbital Insight can rent the data, avoiding the high overhead costs that purchasing imagery would require.
Quantity matters, but so does quality. As Crawford told Earth Imaging Journal, “The resolution of imagery today varies. If you use DigitalGlobe’s WorldView-3 images, you have 30-centimeter resolution and really nice optics.” The WorldView-3 satellite was launched in August 2014 and made 30-centimeter resolution commercially available for the first time. “With a WorldView-3 image” Crawford noted, “you really could tell the difference between a car, a van and a truck.”
Orbital Insight then integrates DigitalGlobe’s geospatial imagery with other libraries of data for applications in commerce, growth, construction, farming, deforestation, poverty, and more. They can then run their own algorithms or use one of ours for information extraction from imagery data sets at scale. Imagine what you could do.
CLICK HERE to download the full white paper on how Orbital Insight leverages Geospatial Big Data through DigitalGlobe’s GBDX platform.
Are you interested in learning more about GBDX? Contact us today!
AdministratorTHE OTIUM POST