Murphy's Law: Roboanalyst Saves The Day

Archives

February 21, 2015: The U.S. Air Force has released some of the statistics on its UAV operations. No real surprises, just confirmation that the use of UAVs was very useful but brought with it some new problems. One major and unexpected difficult was data overload. In 2012 the UAVs were collecting over 1.3 terabytes of data (most of it digital video) each day. By 2012 the air force had already been hustling for several years to find (or develop) software that could assist in analyzing all this data quickly. The air force already knew that they could not recruit and train (much less afford) enough human analysts to do the job. New analysis software showed promise and was getting better each year, but was still largely inferior to a skilled human analyst. The problem was only going to get worse because of the growing number of UAVs and aircraft creating these digital images.

By 2012 the air force had about 400 aircraft (over half of them UAVs) collecting images (digital video and stills) and electronic communications. What the UAV users desperately needed was more powerful pattern analysis software that was good enough to understand what digital video had recorded UAV users wanted software that worked with pattern recognition software to instantly monitor video images and alert human controllers if something of interest was spotted below and keep looking for certain images until ordered to move on by the human controller. The air force kept falling behind in making the most of all this data despite the fact that in 2012 the air force had 20,000 people assigned to maintaining and operating those 400 aircraft and analyzing the data collected. The air force knew this because in cases where they did provide enough analysis capability (software and human) the results were outstanding.

Since 2012 it has become apparent (from the appearance of better digital photo analysis software for civilian security use) that the air force has made more progress in coping with the flood of data. It also appears that air force use of these aircraft for surveillance has been producing more “actionable targets” (something worth firing a missile at). The air force is not the only, or even the principal, customer for this analysis software. Security firms had been developing software like this for civilian security cameras, because it was very expensive to properly monitor all the security video data using humans. The military had a similar problem which required the analysis software to be modified. As with civilian users the military also wanted to store digital video in a database and analyze it again, and again, as intelligence came up with something new to look for. Much has been spent storing these digital images because it is already apparent that this pays big dividends in finding patterns the enemy had fallen into that could be exploited. Naturally, all this stuff is classified, but based on past development of military technology the same patterns appears to be at work with UAV digital imagery and other electronic data collected.