A comparison of dynamic programming tracking methods
C.R. Walters · 1995
This paper considers a number of techniques which use a dynamic programming (DP) search method to detect target tracks. The impetus for this work arises from research into frequency tracking based upon a graph theoretic approach which necessitates a search through all possible potential target trajectories. An efficient implementation of this search can be performed using DP with a cost-function based upon signal-to-noise ratio. During this work, the similarity of this method to a number of other tracking techniques employing DP became apparent and was investigated further, the results being described. Note that although the context of the work is that of frequency tracking, the concepts can be extended to positional tracking without difficulty. The paper considers DP tracking methods generally and focuses onto two techniques, the graph-theoretic (GT) tracker and a hidden Markov model (HMM) tracker, comparing their performance using simulation. (5 pages)