Performance analysis of a track before detect dynamic programming algorithm

Abstract

"track-before-detect" (TBD) is a target tracking technique where the data is processed over a number of frames before decisions on target existence are made. The aim of this paper is to use extreme value theory to analyse the performance of a dynamic programming based TBD algorithms. Asymptotic expressions are obtained for the false alarm and track detection probabilities using extremal analysis of limiting distributions. Apart from fitting the simulated results far more accurately than previous works in the TBD literature, our analysis does not require the unrealistic assumptions of independence and Gaussianity.