Bayesian Characterization and Detection of Rare Binary Signature Events

摘要

A method of imposing a binary classification on a target signature is describedthat transforms a sequence of signatures into a binary-valued time series using the theory of runs. Under the assumption that the time series can be treated as a set of Bernoulli trials, a Bayesian method of estimating a probability density characterizing the outcome of each trial is considered. The density is used to detect signature events which are found to be rare with respect to the classification imposed on the signatures. Finally, tests of homogeneity are used to partition the observed signatures, when necessary, into equivalence classes having the same density characterization. Keywords: Signature analysis, Rare event detection, Change detection, High earth orbiting targets, Ground based radar, Statistical analysis, High altitude targets. (kr)