NVision: Leveraging Artificial Intelligence and Machine Learning to Detect Threats

The proposed NVision system addresses the compelling need of the Air Force and the armed forces generally for operating base force protection to safeguard lives and resources from conventional and asymmetric threats. Most expeditionary and fixed bases have motion detection, visible and infrared video sensors, and sensor zones for intrusion detection. Threats can be complicated events operating in multiple domains. While the proliferation of security sensors provides raw data, human operators must generally assess the sensor observables and quickly decide if a threat is detected. Operators are overloaded by the volume of data feeds and may not recognize the threat or fail to recognize it in a timely manner. Machine Learning (ML) has the potential to automate the detection and characterization of threats, provide a threat assessment, and reduce the workload on force protection video surveillance operators. The benefits realized with such automation include faster, more accurate threat detection, improved courses of action, and reduced cost. In Phase I, a proof of concept capability was developed that demonstrated the potential benefits of applying ML to sensor feeds for force protection. In Phase II, we mature and extend this ML capability to provide a useable prototype.