Washington D.C. gets AI-enabled air defense camera upgrade — new system replaces 22-year-old cameras installed after 9/11
It's like upgrading to an iPhone 16 from an iPhone 4S.
The U.S. Department of Defense (DOD) just deployed a camera system that uses an artificial intelligence-based visual recognition and identification system called the Enhanced Regional Situational Awareness (ERSA) system. The DOD says that the ERSA system is closely monitored by the Eastern Air Defense Sector (EADS), based in Rome, New York and some 300 miles north of the White House. If the Joint Air Defense Operations Center (JADOC) based in Joint Base Anacostia-Bolling cannot determine the identity of radar contact, it will use the AI-equipped cameras to visually check the approaching bogey.
“If we need to validate some radar data that we can’t for sure say what it is, we can utilize the camera system as an asset to look in that set location to assist in the validation process,” says Air Force Master Sgt. Kendrick Wilburn, the non-commissioned officer-in-charge of capabilities and requirements at JADOC. If the operators of the ERSA determine that the unidentified aerial vehicle is a threat, it could then warn the unauthorized traffic to change direction and get out of the Washington metropolitan’s Special Flight Rules Area.
These AI-capable cameras are equipped with electro-optical systems for visual identification, as well as infrared cameras for use at night and low-visibility situations. It also has an eye-safe laser range finder and a warning system that illuminates the cockpit of erring aircraft with red and green lasers. These are more affordable options for interdicting errant aircraft compared to sending two F-16s up in the air for an intercept mission.
Aside from those features, the ERSA system also has machine learning capabilities, making it much more capable at tracking objects with its various locking modes. “The system itself tries to identify what it believes the target to be, and then the operator can assess whether to override it or fine-tune it,” says Wilburn. He also adds, “The camera itself is amazing. We were able to acquire small targets such as a bird flying all kinds of patterns. It locked onto it and held that lock. With the legacies [cameras], you’d have a harder time getting the system to do that.”
Washington, D.C. already has a camera system used for monitoring its air space. It first received this in 2001 after 9/11 and received an upgrade the following year. However, this system stayed in place until the JADOC deployed the ERSA system in 2024. One of the project managers, Marine Corps Maj. Nicholas Ksiazek, likened the upgrade to going from the camera quality of an iPhone 4S to the iPhone 16. Aside from that, the tracking system will make it easier for the operators to follow a tracked object, giving them more time to assess its intentions instead of trying to hold the camera steady on the airplane.
This isn’t the first instance that a military service is using AI in its operations. In fact, the Pentagon has been experimenting with AI technology for a few years now, while Chinese researchers have used Meta’s Llama AI model to build an intelligence chatbot. Let’s just hope that our military leaders do not rely solely on AI for crucial decisions, especially as Jensen himself says that the solution to the AI hallucination problem is still several years away.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Jowi Morales is a tech enthusiast with years of experience working in the industry. He’s been writing with several tech publications since 2021, where he’s been interested in tech hardware and consumer electronics.
-
bit_user
This has nothing to do with that system. It's using AI only for object classification (e.g. bird, plane, drone, kite, etc.) and perhaps to aid in target tracking (i.e. keeping the object centered), which is a feature you can now find even in commodity video surveillance cameras. From the sound of it, the identity of all objects is being visually confirmed by humans, before they decide on any action to take.The article said:while Chinese researchers have used Meta’s Llama AI model to build an intelligence chatbot. Let’s just hope that our military leaders do not rely solely on AI for crucial decisions, especially as Jensen himself says that the solution to the AI hallucination problem is still several years away.
Again, from the article:
"the tracking system will make it easier for the operators to follow a tracked object, giving them more time to assess its intentions instead of trying to hold the camera steady on the airplane."
BTW, I thought the idea of illuminating the cockpit sounded pretty interesting, too. Perhaps AI is also used to find the cockpit windows and keep the laser trained on them. -
DS426 I appreciate the explanation provided by the NCOIC of Capabilities and Requirements of JADOC for this ERSA as closely monitored by the EADS and thanks to the commissioning by the DOD. Oh and AI, just because.Reply
L - 0 - L :ROFLMAO:
Seriously though, looks fancy, is hopefully effective, and probably costs as much as a hypercar. -
bit_user
Oh, more than that, I'd say. Given the low unit volumes and amount of development + testing time needed, it could be much more. I'd be curious to know where else they're deploying such systems, though I assume we'll never find out.DS426 said:Seriously though, looks fancy, is hopefully effective, and probably costs as much as a hypercar. -
williamcll Didn't they found out a week or two ago you can avoid AI cameras by walking in a box Metal gear style?Reply -
bit_user
Anything that makes you look different than what AI cameras know how to recognize can probably help you avoid them. That isn't really news.williamcll said:Didn't they found out a week or two ago you can avoid AI cameras by walking in a box Metal gear style?
In this case, it sounds like the cameras are coupled to a radar-based detection system. The main purpose of camera is to zoom in on the object, so that the operators can get visual confirmation of what it is. AI can help them by suggesting what it thinks the object is, but it's really up to the operator to make the ultimate call.
So, the only way you're going to avoid detection is by using a stealth craft that doesn't even get picked up by the radar. -
SilverStar Heggisist
And this is likely specifically geared towards that. Stealth aircraft are not invisible to radar, show up as small bogies that are usually dismissed by the radar's filter. I read an article about India scrambling fighters over what turned out to be birds that they saw on radar and thought were stealth aircraft.bit_user said:Anything that makes you look different than what AI cameras know how to recognize can probably help you avoid them. That isn't really news.
In this case, it sounds like the cameras are coupled to a radar-based detection system. The main purpose of camera is to zoom in on the object, so that the operators can get visual confirmation of what it is. AI can help them by suggesting what it thinks the object is, but it's really up to the operator to make the ultimate call.
So, the only way you're going to avoid detection is by using a stealth craft that doesn't even get picked up by the radar.
It would be expensive to keep launching fighters to intercept birds.
Similarly it could be used to zoom in on drones. A F16 is gonna do squat against a quad copter and before going all crazy it might make sense to see if that lil drone is a nano that's likely a tourist taking pictures of the city or a terrorists carrying a grenade or something under it. -
USAFRet
Well, the whole of the DC area is restricted airspace.SilverStar Heggisist said:Similarly it could be used to zoom in on drones. A F16 is gonna do squat against a quad copter and before going all crazy it might make sense to see if that lil drone is a nano that's likely a tourist taking pictures of the city or a terrorists carrying a grenade or something under it.
https://airaware.aloft.ai/?lat=38.870572146558686&long=-77.02175540910791
UAS flying is severely restricted. If you have a DJI, the geofencing won't even let it leave the ground without express permission.