A trajectory-oriented Possion Multi-Bernoulli mixture method for matched field tracking to achieve trajectory continuity

Matched field processing (MFP) is a widely used array processing method for target localization. However, weak correlation and low SNR data degrade its performance. Furthermore, submerged targets do not radiate signals all the time for the sake of concealment, which be called as "flickering targets" in this paper. MFP is invalid when the flickering target is silent. In order to achieve trajectory continuity of the flickering target and correct localization errors of MFP, we propose a trajectory-oriented Poisson Multi-Bernoulli mixture method for matched field tracking (TPMBM-MFT). This method substitutes the trajectory state model for the traditional target state model and uses the Poisson Multi-Bernoulli mixture of random finite set (PMBM-RFS) to describe the trajectory state model and the MFT measurement model. The performance is verified by using the SWellEx-96 data set and comparing with MFP localization results and other two tracking methods.

I. Introduction

Matched field processing(MFP) refers to the most widely applied array processing algorithm which exploits the full field structure of signals propagating in an ocean waveguide[1]. Some factors, such as the sound coherence decrease, low SNR of data and environment mismatch, degrade the performance of MFP. Furthermore, the submerged targets do not radiate signals all the time for the sake of concealment, known as the flickering targets. MFP is invalid when the flickering target does not sound. Most of underwater non-cooperative targets belong to flickering targets. Using the measured time series, the data association hypothesis and the prior motion model of a flickering target, tracking methods have ability of inferring the position of a flickering target at silent moments and correcting localization errors of MFP. And then, in order to obtain the continuous position information of a flickering underwater target, the matched-field tracking(MFT) method is extended by including source motion in the parameter landscape of MFP.