Intelligence: Real Time Video Analysis


June 28, 2017: Often high-tech doesn’t work right away, or takes longer than expected to perform to its full potential. Actually, that happens more often than anyone would like to admit. A current example is a new Israeli aerial surveillance system called SkEye. This is a multicamera system built into a UAV as small as the Hermes 450 (max payload 150 kg) or even small manned aircraft like the single engine Cessna 208B. SkEye uses ten high-resolution video cameras to simultaneously watch, track and record activity in an areas as large as 80 square kilometers at a time. SkEye uses digital image scanning software to seek out specific items and patterns of movement below. Since SkEye records all that it sees it can also patterns over time. This approach to aerial surveillance has existed in theory for decades but the technology to make it work didn’t exist until after the 1990s when the U.S. Army needed an effective way to find newly placed roadside bombs and landmines in Iraq and Afghanistan. The Americans spent billions on this bomb detection program and in 2005 the first workable, although somewhat crude, system (“Constant Hawk”) showed up and worked so well that the army named Constant Hawk one of the top ten inventions for 2006. The army does this to give some of the more obscure, yet very valuable, developments some well-deserved recognition. Constant Hawk, like most new technology did not get a lot of media attention. Mainly it's the math, and TV audiences that get uneasy watching a geek trying to explain this stuff in something resembling English. But it worked, and the troops wanted more of it. Tools like this were popular mainly because the system retains photos of areas troops have patrolled, and allows them to retrieve photos of a particular place on a particular day. Often, the troops returning from, or going out on a patrol, can use the pattern analysis skills we all have, to spot something suspicious, or potentially so.

Pattern analysis is one of the fundamental tools Operations Research (OR) practitioners have been using since World War II (when the newly developed field of OR got its first big workout). Pattern analysis is widely used on Wall Street, by engineers, law enforcement, marketing specialists, medical researchers and now, the military. Constant Hawk uses a special video camera system to observe a locality and find useful patterns of changing behavior. While many of the Constant Hawk systems were mounted on light aircraft, others are mounted on towers or other ground structures. Special software compares photos from different times. When changes are noted, they are checked more closely, which has resulted in the early detection of thousands of roadside bombs and terrorist ambushes. This largely eliminated roadside bomb attacks on some supply convoys, which travel the same routes all the time. Those routes were regularly watched by Constant Hawk. No matter what the enemy did, the Hawk would notice. Typically the enemy would shift their bombing efforts elsewhere.

Even as Constant Hawk entered service it was realized that such a system would be even more effective if there was a multiple camera system in a UAV or light aircraft that could detect useful patterns in real time. By 2010 there were two systems that attempted to do that; "Gorgon Stare" and “Angel Fire.” Both were similar to SkEye in many ways but not as compact, capable and reliable. By 2010 there was not as much urgency for the Americans as their troops were being withdrawn from Iraq and were mostly gone from Afghanistan by 2014. But the Israelis still had Islamic terrorists coming at them from several directions and needed better tech to secure its borders and seek out rocket launching sites (all along their Lebanese, Syrian and Egyptian borders) plus roadside bombs and anti-vehicle mines in wartime. Since Israeli firms were responsible for major advances in automated digital image analysis (for automatically checking lots of vidcam images for something specific) and UAV based sensors in general this led to ground based GroundEye in 2016 and the airborne SkEye in 2017.

When the American Gorgon Stare entered service in 2010 it took a year of tinkering to get it to work reliably most (90 percent) of the time. Gorgon Stare used nine cameras (five day and four night/infrared). Aside from enabling several camera operators to work from one UAV, the camera system also has software to enable covering a larger area, by having the cameras cover adjacent areas. The cameras can also look at the same area, from slightly different angles, and produce 3-D images. Two or more cameras can be used over the same area, at different resolution to, for example, search for a specific individual (who is on the Hellfire delivery list) and have another camera focus in on suspect individuals to get a positive identification. The system software also allows for rapidly shifting from one area to another, in response to requests from the ground. Since the RQ-9 UAV that carried Gorgon Stare operates at higher altitudes (7,000 meters or more), the cameras can zero in on particular patches of ground over a wide area.

For a while the Gorgon Stare cameras were providing poor quality images and were unable to stick with people or places operators were trying to keep an eye on. This doesn't mean Gorgon Stare was a failure, it did mean that whoever was responsible for testing the system back in the United States screwed up by overestimating the effectiveness and reliability of the system in a combat situation. Also there was the manpower problem. But even then solutions to both reliability, capability and personnel shortages were in the works. Digital image analysis software had been the next big thing since the 1990s but by 2014 there were some very effective systems in use, especially by the Israelis, for whom getting this right was a matter of life or death.

SkEye was lighter and more compact than Gorgon Stare and with more capable sensors and software provided levels of effectiveness that were largely only theoretical a decade earlier.


Article Archive

Intelligence: Current 2016 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2004 2003 2002 2001 2000 1999