New model cuts through the noise in satellite images
CAMBRIDGE, MA – Decision makers in government, business and philanthropy have long valued information that will help them understand the social, cultural and economic characteristics of a country or region. Information about natural disasters, epidemics, political instability, population growth and agricultural trends can guide decision makers in developing and implementing policies or operations. However, such information is difficult to gather in remote, inaccessible or denied areas. Even orbital satellites, which are growing in number and delivering images at a faster pace, can make it a challenge to distinguish what’s important from the noise.
Draper is addressing this challenge by introducing a new model for understanding satellite imagery through the lens of political, socio-cultural and economic theory and rich sources of survey data. The company, which has designed satellite and space technologies and models and systems to interpret remote sensing data since the 1960s, sees a boom in satellite imagery and the value such images can bring for uses such as optimizing humanitarian aid, capital investment and agricultural planning.
“Thanks to the increase of satellites in orbit, the global community will be able to capture images of nearly every place on Earth, every day by 2017. However, our ability to analyze datasets of these images has not advanced as quickly, until now,” said Kimberly Slater, Program Manager and Earth and Space Science Lead at Draper.
Draper’s new model overlays satellite images with information about a specific area and looks for key indicators. Such indicators might include the relationship between deforestation and economic crisis, and between good governance and social capital as reflected in the presence of infrastructure investments, like schools and places of worship, and infrastructure, such as roads and cell towers.
“Indicators, such as house size, crop health, access to paved roads, and presence of vehicles are indicative of generally higher levels of economic well-being,” said John Irvine, a data scientist at Draper who has published several research papers on satellite imagery. “You can find indicators of economic well-being, governance and social capital through analysis of overhead imagery. In other words, you can model socio-economic phenomena based on imagery observables.”
Irvine’s most recent research revealed a new approach for understanding the indicators of social, economic and governance through satellite imagery. He and his team of researchers compared satellite imagery of three countries in Africa—Botswana, Kenya and Zimbabwe—using both spatial and spectral information to identify key features of interest, such as estimates of the number of buildings, building sizes, building density, extent and health of crops, and types of roads, with attitudinal survey data acquired by the Afrobarometer Program.
The project builds on earlier success and demonstrates broader effectiveness of Draper’s models, which predicted economic, political, and cultural characteristics with an 85 percent success rate using satellite imagery of Afghanistan with funding provided by the Office of Naval Research. The National Geospatial-Intelligence Agency funded the current work on satellite imagery.
In recent years, modeling socio-economic phenomena based on imagery observables has emerged as a promising decision making tool. In the future, governments, businesses and philanthropies might be able to do the same interpretation of satellite images on an off the shelf computer program or smartphone app.
Draper combines specific domain expertise and knowledge of how to apply the latest analytics techniques to extract meaningful information from raw data to better understand complex, dynamic processes. Our system design approach encompasses effective organization and processing of large data sets, automated analysis using algorithms and exploitation of results. To facilitate user interaction with these processed data sets, Draper applies advanced techniques to automate understanding and correlation of patterns in the data. Draper’s expertise encompasses machine learning (including deep learning), information fusion from diverse and heterogeneous data sources, optimized coupling of data acquisition and analysis and novel methods for analysis of imagery and video data.