A Track-Before-Detect Approach to Multi-Target Tracking on Automotive Radar Sensor Data

TL;DR

This paper introduces a novel Track-Before-Detect (TBD) approach using the Generalized Labeled Multi-Bernoulli (GLMB) filter for multi-target tracking in automotive radar data, avoiding information loss and demonstrating real-world applicability.

Contribution

The paper presents the first application of the GLMB filter within a TBD framework for automotive radar data, enhancing multi-target tracking without thresholding.

Findings

Successful application of GLMB filter to real radar data

Improved multi-target tracking accuracy in automotive scenarios

Demonstrated feasibility of TBD GLMB in real-world data

Abstract

In recent years, Bayes filter methods in the labeled random finite set formulation have become increasingly powerful in the multi-target tracking domain. One of the latest outcomes is the Generalized Labeled Multi-Bernoulli (GLMB) filter which allows for stable cardinality and target state estimation as well as target identification in a unified framework. In contrast to the initial context of the GLMB filter, this paper makes use of it in the Track-Before-Detect (TBD) scheme and thus, avoids information loss due to thresholding and other data preprocessing steps. This paper provides a TBD GLMB filter design under the separable likelihood assumption that can be applied to real world scenarios and data in the automotive radar context. Its applicability to real sensor data is demonstrated in an exemplary scenario. To the best of the authors' knowledge, the GLMB filter is applied to real radar data in a TBD framework for the first time.