Press Releases > Advanced Drone-video Overlay Using Machine-learning (ADOM)

Advanced Drone-video Overlay Using Machine-learning (ADOM)

Press Release

Press Release

Physical Sciences Inc. (PSI), has been awarded a research program from the U.S. Army to develop algorithms to overlay Geographic Information System (GIS) and target tracking data on the visible and infrared full-motion video stream from the MQ-1C Gray Eagle.

The ultimate objective is to reduce the mental burden and improve the situational awareness of the payload operator. PSI’s system, the Advanced Drone-operator Overlay using Machine-learning (ADOM), will accurately overlay features and metadata while utilizing state-of-the-art machine learning to perform automated detection of objects of interest. Automated detection will employ custom infrared convolutional neural network (CNN) models previously developed by PSI. The CNN architecture is designed to operate in real-time on embedded hardware, compatible with the resources of the Universal Ground Control Station. These machine learning models will be customized using self-supervised and meta-learning techniques to minimize data requirements and reduce training computational resource requirements. Additionally, ADOM will correlate detected per-pixel features with GIS features to refine the position of the overlay and provide input to Vision Based Navigation systems. These algorithms will draw from PSI’s experience in the development of machine learning and computer vision algorithms, and deploying these algorithms on embedded platforms.

The ADOM framework will reduce the operator burden of unmanned-aircraft systems (UAS) by overlaying key information directly on real-time sensor video on the ground station. ADOM will be directly optimized for use with the MQ-1C Gray Eagle for use by the payload operator, but the developed technology will also improve the capability of PSI’s small-UAS InstantEye® (IE) systems. In both cases, the overlay of GIS information and battlefield metadata improves the operators’ understanding of their relative position to ongoing events. The ability to detect and track relevant targets directly in sensor imagery reduces the possibility of target fixation and loss of situational awareness. InstantEye® systems are used by the military, law enforcement, and first responders for ISR and search and rescue (SAR). Specifically, ADOM will improve the IE system by making vehicle operation more intuitive by providing additional context on the relative position of known GIS features, such as buildings and street names. Additionally, for customers interested in SAR applications, the ability to automatically detect key targets in extremely cluttered scenes would provide a new transformative capability.

For more information, contact:
Dr. John Wright
Area Manager, Advanced Sensing Technologies
jwright@psicorp.com
Physical Sciences Inc.
Telephone: (978) 689-0003