The Every Camera a Biometric Checkpoint (ECBC) moves biometric-based identity dominance further to the edge for near real-time detection of threatening individuals, preventing infiltration by non-authorized personnel, and authenticating authorized personnel. The primary objective of the desired capability is to rapidly determine or verify a personâs identity from live video surveillance by automatically detecting, tracking and submitting high-quality face images to a cloud-based enterprise face recognition and alerting system. The US Army desires an AI/ML-based face detection, face tracking and face search submission edge-device(s) which is capable of monitoring live video surveillance feeds coupled with an enterprise cloud-based watch list management and face recognition matching capability. This presents unique AI/ML challenges, edge device hardware design challenges, and architectural challenges. The architectural challenges are complex, for example: It is not practical to query an enterprise cloud based matching system like DoD ABIS (where watchlists are maintained) for every face. Edge devices processing four frames a second for 10,000 cameras averaging one face per frame would be 40,000 queries per second. Identity analysis and matching will need to be pushed to the edge. Maintaining regularly updated âwhitelistâ facial biometrics for every camera zone requires integration of facial matching systems and access control data systems to keep whitelists current. The security action business rules are unique to each control zone monitored by a camera. A public facing camera at the exterior of a base, a camera at a controlled zone entry checkpoint, and a general surveillance camera inside a controlled zone will all require different system actions as dictated by local security operations. Our initial design address these challenges by 1) leveraging AI/ML technology for face detection, quality analysis, tracking, and matching, 2) pushing whitelist/blacklist face classification and business rules to the edge in our Integrated Face Analysis and Communication Equipment (IFACE) Tier 1 edge device to expedite security actions and reduce face matching query loads, and 3) providing an IFACE Server at Tier 2 to manage IFACE configurations and rules, manage transactions with Tier 3 Cloud based matching systems and current identity operations and intel systems. The result is a feasible, cost practical, and use flexible system capable of meeting ECBC n
