Geospatial intelligence gets SMART
The Intelligence Advanced Research Projects Activity wants a fully automated way to fuse and process the images and spectral data coming from space satellites so that it can spot changes across the Earth's surface in real time.
Monitoring and analyzing changes across the Earth's surface as they happen may seem like pie in the sky, but the intelligence community's research arm, which specializes in high-risk initiatives, has a plan that brings that challenge down to earth.
The Intelligence Advanced Research Projects Activity wants a fully automated way to fuse and process the images and spectral data coming from space satellites and airborne sensors so that it can spot anthropogenic changes, such as building and highway construction, in real time.
IARPA's Space-based Machine Automated Recognition Technique (SMART) program aims to automate broad-area search, monitoring and analysis of anthropogenic activities using imagery and sensor data merged from a range of sources, including U.S., commercial and foreign satellites. The program, IARPA said in an April 12 draft broad agency announcement, "will demonstrate that [geospatial intelligence] gleaned through data fusion as a whole is greater than the simple sum of GEOINT gleaned from multiple images analyzed in the absence of other imagery, by reducing uncertainties in single-modality data analysis and increasing confidence for assessment of activities."
SMART will also address limitations of traditional broad-area searches -- infrequent satellite or airborne sensor flyovers and environmental factors, such as clouds, that obscure the images. Additionally, the volume of data from commercial and government sources is outpacing geospatial intelligence analysts' ability to manage it, so SMART will use high-performance computing, advanced data analytics as well as artificial intelligence and machine learning to fill in gaps in expertise, infrastructure and data preparation.
The four-year SMART program will focus on five areas:
- Integration of U.S. and foreign government and commercial space-based imagery to detect, characterize and monitor relevant anthropogenic processes.
- Use of machine learning tools to improve understanding of observed activities from sources with varying spatial scales and frequencies.
- Improvements in change detection accuracy and tests as to whether spectro-radiometric observations can be used to detect anthropogenic activity.
- Use of DevOps, networking and cloud solutions to quickly integrate and deliver results of SMART research.
- Reduction of uncertainties in the data to increase the viability of detecting changes in archived and current space-based observations.
Read the draft BAA here.