Military researchers eye electro-optical sensors with machine learning algorithms for defense applications

Oct. 7, 2020
DARPA FENCE seeks to develop a low-latency, low-power, event-based camera and a new class of digital signal processing and machine learning algorithms.

ARLINGTON, Va. – U.S. military researchers are asking industry to develop a new kind of camera and digital signal processing to enable intelligent electro-optical sensors for tactical military applications.

Officials of the U.S. Defense Advanced Research Projects Agency (DARPA) in Arlington, Va., issued a broad agency announcement on Friday (HR001121S0001) for the Fast Event-based Neuromorphic Camera and Electronics (FENCE) program.

DARPA FENCE seeks to develop and demonstrate a low-latency, low-power, event-based camera and a new class of digital signal processing and machine learning algorithms that use combined spatial and temporal information to enable intelligent sensors for tactical military applications.

Neuromorphic describes silicon circuits that mimic brain operation; it exhibits low latency, sparse output, and extreme energy efficiency. Neuromorphic cameras offer sparse output, and respond only to changes in the scene, with accompanying low latency and low power for small-format cameras in sparse scenes.

Related: IARPA kicks-off satellite electro-optical sensors project to detect and characterize events over time

Event-based imaging sensors operate asynchronously, and only transmit data from pixels that have changed, so they produce 100 times less data in sparse scenes than traditional focal plane arrays (FPAs). This leads to 100x lower latency at 100x lower power.

Despite their inherent advantages, existing event-based cameras are not compatible with military applications because military images are cluttered and dynamic. The FENCE program seeks to develop an integrated event-based infrared focal plan array with embedded processing to overcome these challenges.

The FENCE program's primary focus is on developing an asynchronous read-out integrated circuit (ROIC) capable of very low latency and power operation, and a new, low-latency event-based infrared sensor with in-pixel processing.

Related: Electro-optical sensors for multispectral aerial reconnaissance with deep-learning introduced by Elbit

The project also will develop a low-power processing layer that integrates with the ROIC to identify relevant spatial and temporal signals. The ROIC and the processing layer together will enable an integrated FENCE sensor that can operate on less power than 1.5 Watts.

The FENCE program will last for four years. Contractors must have personnel with collateral SECRET clearances and access to an accredited facility and secure communications to support classified development.

DARPA researchers are not interested in proposals that produce spiking event-driven cameras that are not cryogenically cooled or have cutoff wavelengths less than 3 microns; that use low technology readiness level (TRL) detector materials not fielded in military systems; or have amalgamations of existing imagers that are neuromorphic but that are not using the event-driven asynchronous methodology.

Related: Electro-optical sensors key to missile defense

DARPA officials say they plan to award contracts to several companies. Companies interested should submit unclassified abstracts no later than 1 Nov. 2020, and full proposals no later than 16 Dec. 2020 to the DARPA BAA website at https://baa.darpa.mil.

Classified abstracts and proposals should be submitted by post or by messenger to DARPA, attn. Program Security Officer, MTO, 675 North Randolph St., Arlington, VA 22203-2114.

Email questions or concerns to Whitney Mason, the FENCE program manager, at [email protected]. More information is online at https://beta.sam.gov/opp/dd33f3e8d0104290a42f74d8a07e99a3/view.

Voice your opinion!

To join the conversation, and become an exclusive member of Military Aerospace, create an account today!