Satellites have been around for a long time. Ever since the Soviet Union launched the world’s first satellite, Sputnik, in 1957, dramatic advancements in satellite technology have increased governments’ ability to survey the earth in exquisite detail, communicate, and, of course, conduct reconnaissance.
Yet, though satellites have become dramatically more sophisticated and capable since Sputnik, analysis of imagery produced by satellites—which is primarily done by humans—hasn’t advanced as quickly as necessary.
Specifically, with the explosion in the volume of imagery produced by satellites, there simply isn’t enough manpower in the Intelligence Community (IC) necessary to review and analyze the volume of data being produced.
This problem is not solved by existing automated imagery analysis algorithms, which are typically developed to analyze data from a single source. Using data or images from multiple sources requires manual comparison of various data streams and often fails to include and contextualize historic data or data over time. This effectively limits analysts’ ability to provide holistically accurate information to decision-makers.
To help mitigate this problem, the Intelligence Advanced Research Projects Activity (IARPA) launched the Space-based Machine Automated Recognition Technique (SMART) program. SMART, a multi-year program, seeks to use machine-learning to automate broad-area search of satellite imagery to detect, monitor, analyze, and describe the progression of man-made and natural change such as heavy construction or crop growth.
“Given the sheer size of the earth and the complexity and number of man-made and natural events, it’s impossible for analysts to review, digest, and describe every occurrence for policy-makers,” said SMART program manager, Dr. Jack Cooper. “Automating broad-area search of massive areas on earth won’t ever be perfect with algorithms, but if SMART can effectively find and keep track of what’s changing then it will free up analysts to dig deeper on their expert analysis of high priority issues, which is what humans are best at.”
Initially, the SMART program will detect and monitor heavy construction projects and, based on these results, subsequently determine the feasibility of transferring this approach to other man-made and natural phenomena. If successful, this will validate the efficacy and transferability of SMART technology.
To bring SMART to fruition, IARPA has contracted with firms (performers) that have deep experience with the development of satellite technology. This includes Accenture Federal Services, Applied Research Associates, Astra, BlackSky Geospatial Solutions, Intelligent Automation Inc, Kitware, and Systems Technology Research.
“We’re really excited about the potential these new technologies have to assist with imagery analysis,” Dr. Cooper said. “If our performers can prove SMART works, I really think it’ll be a game-changer for the IC.”
