Graphical, Optimization, and Learning Methods for Fusion and Exploitation in Sensing and Surveillance Systems.

摘要

This report summarizes our accomplishments under this grant. Our objective is to carry out fundamental research in several interrelated areas: (a) development and use of graphical and hierarchical representations for complex phenomena and for the construction of scalable algorithms for the fusion of heterogeneous sources of information; (b) development of first principles methods for constructing statistical models for the variability of shapes and configurations of objects of interest for statistically optimal shape estimation and object recognition; and (c) development of new adaptive learning and optimization algorithms for analysis of complex, multimodal data for the linking and fusing disparate sources of information, for the characterization of features in complex data and imagery, and for sensor resource management. Our research blends methods from statistics and probabilistic modeling, signal and image processing, optimization, mathematical physics, graphical models, and machine learning theory, yielding new approaches to challenging problems in sensing and surveillance. Moreover, each aspect of our research is directly relevant to Air Force missions. In all of these areas we have contacts and interactions with AFRL staff and with industry involved in Air Force programs.