Enhanced Neural Network for Target Recognition Improvement System

To address the Army’s need for advanced target recognition algorithms in gun-launched munitions, Intellisense Systems, Inc. (Intellisense) proposes to complete the development of a new Enhanced Neural Network for Target Recognition Improvement (ENNTRI) system, based on deep neural network (DNN)-based sensor image enhancement, fusion, and autonomous target recognition (ATR). Specifically, the innovation in its flexible and computationally efficient DNN-based automatic target enhancement and recognition framework and its implementation on commercial off-the-shelf (COTS) accelerated computing hardware (such as graphical processing units (GPUs) and field-programmable gate arrays (FPGAs)) will enable the ENNTRI system to overcome the resolution limitations of onboard sensors and improve the performance of target recognition in terms of accuracy. As a result, the ENNTRI system offers real-time target recognition capabilities in a compact, power-efficient, hardware-accelerated, and ruggedized package suitable to 155 mm artillery size, weight, power, cost, shock, and temperature constraints and operational requirements (extended slant ranges and large field of view). In Phase I, Intellisense developed and prototyped ENNTRI algorithms on representative hardware and tested and demonstrated improved target discrimination capabilities. In Phase II, Intellisense plans to advance the ENNTRI system to a refined prototype system able to perform automatic resolution enhancement, sensor fusion, and ATR, with focus on refinement of the overall ENNTRI system design, optimization of ENNTRI processing algorithms, integration of the ENNTRI system with a representative gun-fired munition, and testing at government test ranges to demonstrate its ability to discern multiple disparate targets within the timing required in multiple environmental conditions.