Oct 11, 2017 -- Members of U.S. defense and intelligence agencies are drowning in vast amounts of high-resolution imagery they need to analyze every day to monitor events unfolding around the world. There is simply not enough manpower to effectively analyze all the image data collected today, and the problem is only getting worse. As a result, defense agencies are looking to utilize machine learning to help automate repetitive and time-consuming image analysis tasks; thereby freeing up the workforce to spend more time focused on harder intelligence problems.
In a study released today, researchers from the Center for Geospatial Intelligence (CGI) at the University of Missouri (MU) demonstrated how machine learning methods could be used to help with the labor-intensive process of human visual searches for military features of interest in large volumes of high-resolution satellite imagery.
Specifically, they used a deep learning neural network to assist human analysts in visual searches for Surface-to-Air Missile (SAM) sites over a large study area in southeastern China. The results showed that the deep learning approach had an average search time of only 42 minutes for the search area of approximately 90,000 km2. This was more than 80 times more efficient than a traditional human visual search, which had an average visual search time of 60 hours, while achieving the same overall statistical accuracy (90 percent) for correctly locating the missile sites.
"We wanted to test these deep learning methods on a realistic, real-world image analysis problem to critically assess their utility and potential impact," said Curt Davis, director of the CGI and professor of electrical engineering and computer science at the MU College of Engineering. "The results were much better than we anticipated. Historically machine learning algorithms haven't performed well when they have been applied to large satellite imagery datasets."
The research study was published in the SPIE Journal of Applied Remote Sensing in a special issue on Deep Learning in Remote Sensing Applications. You can read the full study here. You can search for Chinese SAM sites on your own at this demonstration website that uses the same high-resolution satellite imagery and deep learning algorithms used in the study.
_____________________________
About the MU Center for Geospatial Intelligence (CGI): The CGI was founded in the MU College of Engineering in 2004 with a primary mission to engage in research and development efforts in support of U.S. defense and intelligence agency objectives and related industry. The CGI is an interdisciplinary research center that with faculty and students from Electrical Engineering, Computer Engineering, Computer Science, Geography, Civil & Environmental Engineering and Geological Sciences. The CGI conducts interdisciplinary research in a wide variety of areas including, but not limited to: satellite, airborne, and ground-based remote sensing; high performance computing for geospatial data analytics; automated feature extraction; change detection; pattern recognition; machine learning; video processing and surveillance; and human geography and modeling. For inquiries, please contact us at Email Contact or 573-268-4908.
Contacts:
Curt H. Davis, Ph.D., Fellow IEEE
Naka Endowed Professor
Director - Center for Geospatial Intelligence
Email:
Email Contact
Learn more about
F. Robert Naka
W1025 Lafferre Hall
Phone: 573-884-3789
College of Engineering
Fax: 573-884-1626
University of Missouri
Cell: 573-268-4908
Columbia, MO 65211
