Topological and Statistical Behavior Classifiers for Tracking\n Applications

Abstract

We introduce the first unified theory for target tracking using Multiple\nHypothesis Tracking, Topological Data Analysis, and machine learning. Our\nstring of innovations are 1) robust topological features are used to encode\nbehavioral information, 2) statistical models are fitted to distributions over\nthese topological features, and 3) the target type classification methods of\nWigren and Bar Shalom et al. are employed to exploit the resulting likelihoods\nfor topological features inside of the tracking procedure. To demonstrate the\nefficacy of our approach, we test our procedure on synthetic vehicular data\ngenerated by the Simulation of Urban Mobility package.\n