Deep learning research by NSWC Crane engineer to help electronic warfare capabilities
NewsJuly 11, 2019
CRANE, Indiana. ? Research on how to leverage artificial intelligence (AI) for electronic warfare applications was completed by a Naval Surface Warfare Center, Crane Division (NSWC Crane) employee.
David Emerson, a NSWC Crane engineer at defended his research that make use of deep learning to process images and determine distances between objects in a scene. Self-driving cars are an example of such a system as they use deep learning to reconstruct scenes in 3-D, but Emerson explains there is interference that influences the effectiveness of currently used methodology. He says depth estimation is one of the most challenging problems in computer vision.
“The human eye can see and understand depth in a scene,” Emerson explains. “A computer sees 2-D images and has to calculate the distance to reach an ‘understanding’ of depth between objects. In my Depth from Defocus method, I took one photo in-focus and one photo out-of-focus and constructed a 3-D image.”
Emerson’s Depth from Defocus (DfD) using deep learning (DL) methodology out-performed previously used methods, increasing the speed of the process, and was more robust in its handling of images in low-lighting conditions, according to a NSWC Crane release.
“In the military, technology is used to determine long distances in the field,” he says. “DfD and deep learning methodology has potential future applications that could considerably improve the warfighter’s speed and capability, all while remaining stealthy.”
Emerson’s research was partially funded first with a Department of Defense (DoD) Science, Mathematics And Research for Transformation (SMART) Scholarship for Service Program, and then later with full sponsorship by the NSWC Crane Ph.D. Fellowship Program.
“I would not have been able to pursue this research if it weren’t for the flexibility and support of my manager, Christopher Crombar, and team,” Emerson says. “Overall, the learning process has been beneficial. Machine learning and deep earning are not going to go away; researching the latest trends now can translate to solving future problems for military, industry, and academia. I am excited to apply everything I’ve learned over the past few years to help people and save lives.”
