Multi-Object Tracking with an Adaptive GLMB Filter

This is an implementation of the adaptive GLMB filter proposed in:

@ARTICLE{adaptive_GLMB, 
  author={Do, Cong-Thanh and Nguyen, Tran Thien Dat and Moratuwage, Diluka and Shim, Changbeom and Chung, Yon Dohn}, 
  journal={Signal Processing}, 
  title={Multi-object tracking with an adaptive generalized labeled multi-Bernoulli filter}, 
  year={2022},
  volume={196},
  pages={108532}}

The paper is available at https://www.sciencedirect.com/science/article/abs/pii/S0165168422000792
A pre-print version is available at https://arxiv.org/abs/2008.00413

Implementation Notes

'v1' is the original filter proposed in the paper.
'v2' is based on the filtering algorithm proposed in [1].
Schematic of the 'v2' filter

In 'v2' implementation, the detection probability for each track is processed by GLMB filter.

Run Instructions

Use the file 'demo.m' to run the demonstrations. In this file:

• Line 67 runs demonstration with linear Gaussian models and filter implementation 'v1'.
• Line 68 runs demonstration with linear Gaussian models and filter implementation 'v2'.
• Line 69 runs demonstration with non-linear Gaussian models and filter implementation 'v1'.
• Line 70 runs demonstration with non-linear Gaussian models and filter implementation 'v2'.

You can choose to use either the standard estimator or partial smooth estimator proposed in [2] by setting the parameter 'filter.estimator_type' in files 'gms/run_filter_v1.m', 'gms/run_filter_v2.m', 'ukf/run_filter_v1.m' and 'ukf/run_filter_v2.m'. Partial smooth estimator is used by default.

Performance Notes

The algorithms occasionally overestimate the cardinality due to the nature of the measurement-driven birth model.

Acknowledgment

This implementation is based on MATLAB RFS tracking toolbox provided by Prof. Ba-Tuong Vo at http://ba-tuong.vo-au.com/codes.html.

References

[1] C.-T. Do, T.T.D. Nguyen, and H.V. Nguyen. 2022. "Robust multi-sensor generalized labeled multi-Bernoulli filter" Signal Processing 192, pp. 108368.
https://www.sciencedirect.com/science/article/pii/S0165168421004059
[2] T.T.D. Nguyen, and D.Y. Kim. 2019. "GLMB Tracker with Partial Smoothing" Sensors 19, pp. 4419.
https://www.mdpi.com/1424-8220/19/20/4419

Contact

For any queries please contact me at tranthiendat.nguyen@gmail.com.
Copyright (C) 2022, Tran Thien Dat Nguyen.