The United States Defense Advanced Research Projects Agency (DARPA) selected three teams of researchers led by Raytheon, BAE Systems and Northrop Grumman to develop infrared (IR) camera technology that responds to the surrounding environment. It is a project within the FENCE program, which aims at a "neuromorphic" electronic camera, with an artificial vision that it mimics the way the brain processes information.
Today's state-of-the-art cameras work well with scenes that have few edits to track and with images that are relatively straightforward. However, their abilities fail in highly messy and dynamic scenes, limiting their use in many military applications.
How the "brain" camera works
DARPA's FENCE program aims to develop a new class of event-based infrared Focal Plane Focal Array (FPA) algorithms, and low-latency, low-power digital signal processing and machine learning algorithms. In other words? The development of a neuromorphic camera will allow intelligent sensors to handle more dynamic scenes for future military applications.
Event-based imaging sensors work asynchronously, and only transmit information about the pixels in an image that have changed. This means that they work at breakneck speeds, on very complex images, but produce much less data, while consuming less power. A hyperbolic leap forward.
The Neuromorphic Camera has silicon circuits that mimic the functioning of the brain; they offer low output, low latency and high energy efficiency. An event-based camera works according to these principles, but today it lacks advanced intelligence to perform complex tasks.
dr. Whitney Mason, program manager leading the FENCE program
The next steps
Researchers from Raytheon, BAE and Northrop they will work to develop an asynchronous readout integrated circuit (ROIC) that will allow an integrated FENCE sensor to operate at less than 1,5W of power.
Under these conditions, believe me, there is nothing that a camera of this type will be able to ignore.
