Networked Enhanced Automated Target Recognition (NEATR), is an important problem in modern network centric operations. This Phase I effort focuses on fusion in an embedded system with reasonable Size Weight and Power (SWAP). The proposed multi-layer approach provides novel detection and re-identification layers, target tracking and enhanced PCA-based recognition. The proposal presents a unique new approach, P+PCA simultaneously estimating, in a distributed multi-hypothesis manner, target position, pose and coefficients in a parametric PCA-eigenspace. Estimation of position and pose are traditional target tracking fusion issues, we integrate them with the estimation the parameters and uncertainties needed to improve the distributed ATR performance from reduced resolution data. The approach is designed for low-bandwidth communications and does not transmit the imagery for fusion. The effort builds on the team's decade of experience delivering on DOD-funded surveillance and recognition R&D efforts. This Phase I will develop a functioning FPGA-based prototype to address key feasibility questions, new algorithms tested on real data and will develop and analyzes the overall architecture for distributed recognition.
Keywords: Imaging Sensors, Distributed Fusion, Situational Awareness, Video Surveillance, Covert Surveillance, Embedded Systems, Automated Target Recognition, Low-Power Sensors
