Particle initialization effect of particle filter based track-before-detect algorithm

Abstract

The problem of detecting and tracking weak, stealthy or dim targets is a challenging one. The particle filter based track before detect (TBD) algorithm is appealing for its generality and potential for solving non-linear non-Gaussian problem. However, real-time implementation can be hardly realized by this class of algorithm. Current state-of-the-art research ubiquitously deal with computationally intensive problem by improving the algorithm itself. This paper presents a recent study on the effect of the initial velocity range and position range of the particles on the existing particle filter based TBD methods to decrease the computational time. Moreover, it points out that a suitable velocity range and position range will improve detection ability and tracking performance.