GLOBAL DEFENSE MARKET
New Algorithm Tackles Big Data Problems
Lockheed’s algorithm, referred to as “quick adjust,” offers an alternative option for “geomosaicking” images taken by small, unmanned aerial vehicles. Geomosaicking is used to describe the process of taking tiny photos captured by a drone and adjusting them so they line up together to form one large image. Compared to “bundle adjustment,” which is the current market competitor, quick adjust theoretically has no limit to the number of images it can process, said Mark Pritt, a systems engineer at Lockheed. Bundle adjustment stops working at around 100 images.
“In the case of 1,000 images, [quick adjust] is 100 times faster,” said Pritt. “Bundle adjustment, as a mathematical technique, just doesn’t do well when you have over 1,000 images.”
The need for a faster imaging algorithm will increase as the problem of big data grows, said Lee Hall, director of exploitation and visualization solutions at Lockheed.
As the Federal Aviation Administration begins to open up the airspace for domestic use of drones and the number of commercial satellite companies launching large constellations of small satellites for Earth observation increases, the management of big data emerges as a significant problem.
So far most of the field-testing for the algorithm has been in crop monitoring on farms, but there is significant potential for defense and first responder applications, Lockheed officials said.
“For defense … whenever the military needs to create an accurate map of an area, this will be perfect for them because it will be a very low cost way to do it,” Pritt said, adding that it will be about 10 times cheaper than the current process of using a manned aircraft. It is ideal for unsafe fly zones where a pilot’s life would be at risk, he said.
There is also potential in border security and crisis response. If a first responder needs to determine how to deploy into an area or how to direct resources, the algorithm would be a good way to collect data to inform those decisions, Hall noted.