Algorithms for Tracking in Clutter and for Sensor Registration

Abstract

In this dissertation, we consider various aspects of tracking multiple targets in clutter. The primary focus is on target tracking algorithms. Specifically, we derive and extend upon variants of the Probabilistic Multiple Hypothesis Tracker (PMHT), variants of the Joint Probabilistic Data Association Filter (JPDAF), and variants of the Multiple Hypothesis Tracker (MHT). However, to be able to track targets, more is needed than just a tracking algorithm. Among other things, the sensors observing the targets must take measurements in a common coordinate system. For that reason, we have considered algorithms for sensor localization and for residual bias estimation. Similarly, since merged measurements can pose problems for target tracking algorithms, we reviewed and rederived an old algorithm for resolving two closely-spaced targets in angle when using a four-channel radar array. Finally, we also considered and extended Gaussian mixture reduction algorithms, which are necessary for the use of certain variants of the MHT that merge hypotheses instead of or in addition to pruning them.