Draper Labs leverages commercial imagery for socio-cultural understanding
Satellite imagery is an untapped resource for understanding human geography, according to John Irvine, the chief data scientist at Draper Laboratory.
The National Geospatial-Intelligence Agency (NGA) is funding research at Draper in which computer models analyze commercial satellite imagery to reveal socio-cultural conditions such as economic wellbeing, community engagement, and attitudes toward governance.
The NGA-funded research, expected to wrap up in April, studies the effectiveness of Draper’s image-based Socio-Cultural Observation, Prediction, and Estimation (iSCOPE) models using imagery of Sub-Saharan African. The results are compared with survey data for validation. iSCOPE is part of an integrated portfolio at Draper intended to help government and industry analysts make better use of Big Data.
The traditional way of learning a society’s human geography is through direct observation, survey data, and media. Draper’s goal is to take advantage of new and growing sources of information in commercial remote sensing to collect information about society more cost effectively, over broader areas, and through a more reproducible process, according to Irvine.
“Lots of people need to be well informed about societies when doing business in different parts of the world—trying to engage in commerce, enact policy decisions, address problems related to health and wellbeing,” Irvine said. “Often there is a shortage of data or if available it is expensive to collect.”
The lab’s current research is intended to build upon previous success and demonstrate broader effectiveness. With prior funding from the Office of Naval Research (ONR), Draper had an 85 percent success rate in predicting socio-cultural characteristics by applying its models to satellite imagery. The imagery results were compared with survey data from 500 villages in northern Afghanistan.
Although survey data is Draper’s standard for truth while it conducts research, the future expectation is the models will be robust enough to be applied in other areas without the need to conduct surveys.
Satellite imagery can illuminate human geography in many ways, according to Irvine. For example, tree cover or the size, spacing, and orientation of houses can reveal the local level of income. The density and connectedness of a village—such as its access to paved roads or railroads—can often shed light on its citizens’ political engagement and attitudes about society as a whole.
The current study of Sub-Saharan Africa focuses on Botswana, Kenya, and Zimbabwe. The region was selected because it represents a cross-section of societal and economic differences from culture to urbanization.
Irvine said SmallSats and the growing number of commercial remote sensing data sets would bolster the ability to leverage commercial imagery for socio-cultural understanding.
“I was thrilled with our initial work for ONR,” Irvine said. “The opportunity to extend that work now with NGA is great. This has the potential to really change the way people approach human geography.”
DigitalGlobe image provided by NGA
Hosted by Military Open Source Software (Mil-OSS) and the United States Special Operations Command (SOCOM) as well as sponsored by USGIF and a number of other organizations, the first Tampa Hackathon was held at SOCOM’s SOFWORX collaboration facility in Tampa, Fla.