A particle belief propagation algorithm based on track-before-detect for tracking low-observable and maneuvering targets using multiple sensor

It is notoriously challenging work to track an unknown number of low-observable maneuvering targets. In this paper, a sequential Bayesian inference method based on multiple-model dynamic model and track-before-detect measurement (TBD) model is proposed for tracking low-observable maneuvering targets using multiple sensors. Our proposed method is based on a new multi-sensor statistical model that allows targets to interact and contribute to more than one data cell for pixeled image TBD approach. Based on the factor graph representing multi-sensor statistical model, the marginal posterior densities are derived by performing the message passing equations of the proposed belief propagation (BP) algorithm for target detection and target state estimation. The simulation results validate that the computational complexity of our proposed multi-sensor BP-TBD algorithm scales in the number of sensor nodes and demonstrate that its performance is superior among the state-of-the-art multi-sensor TBD methods.