A BP Method for Track-Before-Detect

TL;DR

This paper introduces a novel belief propagation-based track-before-detect method that directly utilizes raw sensor data for tracking multiple low-observable objects, improving accuracy and scalability.

Contribution

It presents a new statistical model and BP algorithm for TBD that allows objects to interact with multiple data cells, enhancing tracking performance.

Findings

Outperforms existing TBD methods in simulations

Reduces computational complexity through message approximations

Effectively tracks multiple low-observable objects

Abstract

Tracking an unknown number of low-observable objects is notoriously challenging. This letter proposes a sequential Bayesian estimation method based on the track-before-detect (TBD) approach. In TBD, raw sensor measurements are directly used by the tracking algorithm without any preprocessing. Our proposed method is based on a new statistical model that introduces a new object hypothesis for each data cell of the raw sensor measurements. It allows objects to interact and contribute to more than one data cell. Based on the factor graph representing our statistical model, we derive the message passing equations of the proposed belief propagation (BP) method for TBD. Approximations are applied to certain BP messages to reduce computational complexity and improve scalability. In a simulation experiment, our proposed BP-based TBD method outperforms two other state-of-the-art TBD methods.