Written byChris Bing
A video software platform straight out of “Minority Report” that layers over live cameras, combining facial recognition technologies with an analytics component to decipher activity and behavior, is in the works. And the U.S. intelligence community is interested in getting its hands on it.
Such a technology “would have aided the response to previous incidents such as the Boston Marathon Bombing,” said Charles A. Carithers, a spokesperson for the Office of the Director of National Intelligence. “It is needed today, because the threats from one or more individuals carrying out an attack continue to grow.”
The Intelligence Advanced Research Projects Activity announced a Proposer’s Day last week for a new program called Deep Intermodal Video Analytics, which aims to “produce a common framework and software prototype for activity detection, person/object detection and recognition across a multi-camera network,” the announcement reads. “The impact will be the development of tools for forensic analysis, as well as real-time alerting for user-defined threat scenarios.”
IARPA anticipates issuing a broad agency announcement solicitation based on the feedback of the event.
The move is significant, in part, because of the technology’s current limitations and given what the government’s public interest in this market can do from a financial stand point, helping drive both supply and development in the tech sector.
The first phase of the program will eventually look to gather data from multiple realtime, streaming feeds via indoor and outdoor security cameras. The next phase — described as phase two — will add body camera capabilities and be able to recognize objects “across multiple overlapping and non-overlapping camera viewpoints.” Lastly and perhaps most importantly, the program will require the creation of powerful cloud-based, scalable framework to easily add bandwidth at an on-demand pace.
At the moment, the biggest issue with this sort of technology relates to a reliable analysis of context in real-time — which translates into accurately knowing when a person and/or object appearing on video is a threat based upon a specific action, moment or movement.
Carithers explained that the most “prominent research or development problems [tied to DIVA] are computer vision and activity detection at a fine-grain level.” A false positive in this field, after all — imagine if a person is incorrectly identified as a bomber in a public square — could result in a tragic outcome.
Surveillance tech firm Tygart Technology’s President John Waugaman told Engadget that the phase 1 of IARPA’s project will be within reach in just two years.
IAPRA’s DIVA Proposer’s Day is scheduled for July 12 in the Washington, D.C. area.
To contact the reporter on this story you can send him an email via email@example.com or follow him on Twitter at @Bing_Chris. Subscribe to the Daily Scoop to get all the federal IT news you need in your inbox every morning at fdscp.com/sign-me-on.