Information and Motion Pattern Learning and Analysis Using Neural Techniques

摘要

This work addressed the development and application of neural models for higher-level information fusion at Levels 2+/3 according to the JDL Data Fusion Group Process Model. We explored several new concepts based on insights from neural processing, learning, and representation. Building on some initial prior work under AFOSR sponsorship, we continued investigation of mechanisms to rapidly and incrementally learn models of normal behavior exhibited by moving tracked entities. These models form the basis of anomaly detection (as deviations from normal) and prediction of future behavior. Our approaches require no ground truth or operator input, but the learned models cab be refined through operator feedback if such is available.