Bayesian track-before-detect method based on Gaussian message passing on graph

摘要

This article presents a Bayesian track-before-detect (TBD) method based on Gaussian message passing to detect and track a target in the low SNR scene. Removing the threshold brings great computation demanding to TBD methods. Because of the distributive law and computation consistency, message passing can reduce the calculation load efficiently. Also, as a probabilistic inference algorithm, message passing can perform the TBD method as a Bayesian posterior distribution. Meanwhile, the close form of multivariate Gaussian message passing is derived, with the adoption of Taylor series expansion to treat the nonlinearity in observation equation. The simulation results demonstrate the effectiveness the proposed algorithm.