A new surveillance technology will allow for the immediate, automatic detection of suspicious behaviour in live video feeds.
The research project, led by the IARPA unit at the U.S. Office of National Intelligence, proposes a recognition system dubbed Deep Intermodal Video Analytics (DIVA), which is able to scour through multiple live video streams and pick out potential criminal and terror activities.
Describing the project, IARPA officials noted: ‘The DIVA program will produce a common framework and software prototype for activity detection, person/object detection and recognition across a multi-camera network. The impact will be the development of tools for forensic analysis, as well as real-time alerting for user-defined threat scenarios.’
The technology is designed to pore through video surveillance and other incoming camera footage to search for new threats, as well as to identify individuals and objects already flagged by the authorities as posing a danger.
DIVA’s strength, according to IARPA, is its capacity to recognise activities at multiple levels of ‘granularity.’ The system can detect across three different phases including; ‘primitive’ behaviours such as ‘person getting into a vehicle’, ‘complex’ activities such as ‘person being picked up by vehicle’, and detection across multiple overlapping and non-overlapping camera viewpoints.
These phases, IARPA explains, can take place on video collected from indoor and outdoor CCTV, fixed and rigid motion cameras, as well as on-body devices. Other sensors could also feed information from other areas of the electromagnetic spectrum (i.e. infra-red), not visible to the human eye.
At a Proposers’ Day, on 12th July, the intelligence body will reach out to the rest of the community for collaboration opportunities. It is expecting to receive interest from experts in machine learning, artificial intelligence, detection, recognition, 3D reconstruction from video, super-resolution, statistics, probability and mathematics. It anticipates that academics and private sector companies will also be keen to participate in further research.