Small Arms Neural-Network Automatic Target Classification

To address the OSD’s need for a new set of algorithms to detect, classify, recognize, and identify all potential targets within the engagement range of small-caliber weapon systems, Intellisense Systems, Inc. (ISI) proposes to advance the development of the new Small Arms Neural-Network Automatic Target Classification (SANATA) system that was proven feasible in Phase I. The SANATA technology integrates a number of very recent developments in machine learning and embedded devices for sensor fusion for high-confidence target classification; this innovative integration enables SANATA to directly address the OSD requirements. During Phase I, ISI successfully developed, implemented, tested deep learning algorithms designed to detect people, body armor, and weapons in multispectral (long wave and short wave) infrared imagery. ISI demonstrated real-time SANATA processing using an edge computing platform, as well as successful object detection in imagery degraded by the presence of simulated fog. In Phase II, ISI will develop and construct a SANATA prototype system (including its performance specifications), featuring automatic target recognition and aim-point selection capabilities. This prototype will be integrated with a small caliber weapon system and tested for operation in both virtual and operational environments, day and night, leading to the identification of key technical challenges.