The U.S. Army is testing a commercial, off-the-shelf AI security system at its Blue Grass Army Depot (BGAD) in Kentucky. A release issued by Joint Munitions Command, Public and Congressional Affairs says the system is “a real-time modular physical threat detection solution using artificial intelligence and deep neural learning computer vision to modernize existing electronic security systems”.
The U.S. Army has previously contracted Idemia NSS for computer vision and facial recognition tech, but the release does not specify which machine learning system or platform is being trialed at BGAD – only that it is “basically a software package that integrates with existing CCTV cameras” to detect “intruders, guns, fights, slips and falls, smoke and fire, facial recognition, and behavior anomalies of individuals.” Flags will send an alert to security response forces.
Echoing the language of other law enforcement and military officials, the provost marshal of Joint Munitions Command, Jim Vaughan, calls the AI system “a game changer for security and protection.” Vaughan says “the Army is looking to use current technology to replace and or backstop antiquated security systems” and that AI is “a great tool that is modernizing existing technology into our force protection posture.”
Potential use cases for the system include active shooter scenarios, “intrusion detection and perimeter protection.” In a demo that simulated a real-time active shooter scenario, Vaughan says AI-based weapon detection typically occurred within seconds in nearly every case. Another demo showcased the system’s capabilities for detecting breaches of an established perimeter.
“The system is able to detect from hundreds of meters away when someone walks into the perimeter,” explains Vaughan. “It can detect if it’s an animal, say a deer, or a person, and it shows where the person is at, and gives a description of what they’re doing. The system can also be trained to look for mechanical flying devices and natural flying devices (birds).”
The system is reportedly currently achieving around 97 percent accuracy in testing.
Funding for the project comes from the U.S. Army’s Physical Security Enterprise and Analysis Group.
Article Topics
access control | biometrics | facial recognition | monitoring | person detection | U.S. Army
