FBI issues RFI on drones with real-time facial recognition, license plate readers

The Federal Bureau of Investigations (FBI) has issued a new Request for Information (RFI) seeking AI tools that are capable of analyzing real-time video from unmanned and manned aircraft, signaling a push to expand the bureau’s use of automated aerial surveillance technologies.

The information gathered may shape the FBI’s eventual requirements for advanced AI-enabled aerial surveillance tools.

The RFI was released by the FBI’s Finance and Facilities Division as part of early market research and does not constitute a solicitation for a contract.

According to the RFI, the FBI is evaluating AI and machine-learning systems that can process live video feeds and detect a range of objects, activities, and identities.

The FBI is specifically looking for systems that can identify vehicles, vessels, people, animals, and firearms, and which can perform facial recognition, read license plates, and analyze perimeters with an awareness of movement direction.

The bureau also wants systems that can integrate with Team Awareness Kit (TAK) tools, including the UAS Tool plugin, and that are compatible with electro-optical and infrared sensors.

TAK is a software system that was originally developed for the military that provides a collaborative platform for situational awareness. It allows agencies to share data like locations, weather, and mission critical information in real-time to enhance coordination during emergencies and operations.

The FBI hosts servers that allow other public safety organizations, like the Texas DPS, to use TAK at no cost, and the system supports communication across different equipment and frequencies.

The UAS Tool plugin is a free, government developed software add-on that allows users to control various drones and share their real-time data such as location and video within the TAK ecosystem. It enables a shared, real-time operational picture by allowing users to pilot drones with TAK instead of the native app, and to stream information to other TAK users, making it a cost-effective tool for public safety agencies.

The RFI emphasizes that the AI systems must be capable of running on-premises with optional deployment on Jetson Orin-equipped unmanned aircraft. The technology is expected to meet a minimum Technology Readiness Level of 7, indicating a system prototype demonstrated in an operational environment.

Jetson Orin is a NVIDIA edge AI platform for autonomous machines and robotics featuring the company’s NVIDIA Ampere architecture GPU and deep learning accelerators. It comes in various modules like the high-performance Jetson AGX Orin, the more cost-effective Jetson Orin NX, and the entry-level Jetson Orin Nano, which can achieve up to 67 TOPS of AI performance.

Developer kits allow for the prototyping of advanced AI applications, from robotics to natural language understanding, and are supported by NVIDIA’s AI software stack.

The FBI is seeking responses from original equipment manufacturers only. Companies are asked to describe their technical capabilities, their ability to provide UAS platforms and AI models, their business size, and whether they can provide solutions under GSA Schedule or NASA SEWP contracts.

Respondents must also specify whether their models are based on YOLO frameworks, whether they can decode KLV or Cursor-on-Target metadata from live camera feeds, and whether their models can be trained locally.

A required capabilities table asks vendors to detail the resolution, altitude limits, and detection performance of their systems across categories such as person, animal, vehicle, firearm, license plate, and facial recognition.

For vehicle detection, the FBI notes that systems capable of identifying vehicle type and color would be advantageous. For perimeter detection, systems able to differentiate between ingress and egress are requested.

Responses of up to 15 pages are due by December 30, at 11:59 p.m. Central Time. Questions must be submitted in writing by December 12 and will be answered publicly on SAM.gov. All submissions must be unclassified and become government property.

Article Topics

drones  |  facial recognition  |  FBI  |  license plate readers  |  real-time biometrics  |  RFI  |  video surveillance