Object Recognition in Support of SOF Operations

Current and future operational environments will increasingly require Special Operation Forces (SOF) to be more self-sufficient while operating in contested and politically sensitive regions where situational awareness can be degraded. This project continues Semi-Autonomous Threat Learning Alert System (SATLAS) efforts to integrate artificial intelligence-enabled small unmanned aerial systems into SOF teams to increase situational awareness and survivability. Specifically, we focus on directing prototype development and evaluating the ability of object recognition software to detect and categorize trained entities including weapons, personnel, and vehicles. Collaborating with commercial industries, we conduct simulation and field experiments to measure the ability of the Surveillance, Persistent Observation and Targeting Recognition (SPOTR) object recognition software to meet the technical requirements of the SATLAS project and operational requirements of SOF teams. We evaluate SPOTR based on accuracy, number of entities detected, and range of detection and recommend methods to improve its performance and meet our determined operational requirements. We advance the SATLAS project and set conditions for subsequent student teams to continue these efforts.