Pith. sign in

REVIEW 4 major objections 4 minor 56 references

The PHD/CPHD filter for Multiple Extended Target Tracking with Trajectory Set Theory and Explicit Shape Estimation

T0 review · 4 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper derives closed-form recursions that let PHD/CPHD filters output whole trajectories and explicit ellipse shapes.

desk verdict A worthwhile combination of trajectory PHD/CPHD filtering with decoupled shape estimation, but the printed recursions contain load-bearing math errors that invalidate the closed-form claims as written. read the letter →

arxiv 2504.15040 v2 pith:KV4TGDR2 submitted 2025-04-21 eess.SP

Yuanhao Cheng , Yunhe Cao , Tat-Soon Yeo , Fu Jie , Wei Zhang This is my paper · ORCID
classification eess.SP
keywords multi-targettrackingextendedtargetunresolvablegrouptargetstrajectorysettheoryPHDfilterCPHDellipticalextentestimationGaussianmixtureimplementation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proposes two closed-form Bayesian filters, the TPHD-E and TCPHD-E filters, for tracking multiple extended or group targets whose sensor footprint is an ellipse. It takes the PHD and CPHD filters — random-finite-set trackers that propagate an intensity over targets and, for CPHD, a full cardinality distribution — and adds Trajectory Set Theory so each Gaussian component represents an entire target history. A decoupled shape state carries the ellipse's orientation and two semi-axis lengths, updated through a linearized pseudo-measurement, so shape estimates emerge explicitly rather than inside a random matrix. In simulations and a real traffic-intersection dataset, the paper reports better shape accuracy than a random-matrix baseline and more complete, stable trajectories than a label-based baseline, with the CPHD variant most accurate in target number.

What carries the argument

The load-bearing mechanism is the trajectory Gaussian component $N(\chi;t,\dot{\xi},\dot{\Xi})$, whose block-diagonal mean and covariance store the target state at every time step of the trajectory, propagated by Trajectory Set Theory. The decoupled shape update uses the multiplicative-noise measurement model of [40]: each measurement contributes a pseudo-measurement $Z=\breve{F}\big((z-\bar{z})\otimes(z-\bar{z})\big)$ whose expectation is a linear function of the shape parameters, keeping the ellipse orientation and semi-axis estimates inside the Gaussian-mixture recursion. Sequential updates and a merging scheme with separate kinematic and shape thresholds keep the mixture manageable. The paper notes that building the trajectory covariance by block-diagonal concatenation in Eqs. (56)-(57) is a slight approximation, since it ignores correlations between states at different times.

What would settle it

Apply the GM-TCPHD-E filter to a single, highly elongated target executing a sharp turn with low detection probability, and compare the estimated $\theta,l_1,l_2$ with ground truth over repeated Monte Carlo runs: if the pseudo-measurement linearization is inadequate, orientation error or Gaussian Wasserstein distance will fail to converge even when cardinality is correct. A second check compares the block-diagonal trajectory covariance against a full-scan implementation on two closely spaced, parallel-moving targets; if cross-time correlations matter, the block-diagonal version should show track fragments or elevated switch cost.

Watch

Extended reading notes

Core claim

The central claim is that trajectory PHD/CPHD filtering and a decoupled elliptical shape model can be merged into closed recursions for multiple extended targets. Kinematic state $r$ and shape state $s=[\theta,l_1,l_2]^T$ are propagated independently, and shape information enters through the pseudo-measurement update of [40], so the posterior remains a Gaussian mixture over trajectories. The paper asserts that the resulting GM-TPHD-E and GM-TCPHD-E filters are strictly defined trackers that avoid trajectory switching in close parallel motion, converge to accurate ellipse orientation and axis lengths, and outperform the GM-TPHD-GIW filter in shape error and the GM-LPHD-E filter in trajectory metric, with GM-TCPHD-E the best overall.

Load-bearing premise

The whole scheme depends on the approximation that the ellipse's shape can be updated with a linearized pseudo-measurement and that each trajectory's state history can be stored with a block-diagonal covariance that ignores correlations between different times; the paper concedes this second part is slightly wrong, so everything rests on that error being negligible.

Editorial extensions

If this is right

  • With these recursions, an extended-target PHD/CPHD tracker can output complete trajectories, not just current states, without a separate labeling or network-flow track-construction stage.
  • The ellipse orientation and semi-axis lengths are estimated explicitly in closed form, which is what obstacle avoidance and formation-flight applications need.
  • The CPHD version, GM-TCPHD-E, should estimate the number of targets more accurately at births and deaths, and the paper's experiments show it is least affected by parameter changes.
  • The separate kinematic and shape merging thresholds let the filter keep shape-distinct components alive while pruning kinematically redundant ones.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same decoupled shape update could be combined with measurement-driven birth models, such as Poisson multi-Bernoulli filters, which the authors list as future work, to remove the dependence on a hand-placed birth density.
  • Because the shape merging threshold is set smaller than the kinematic one, two targets with similar kinematics but distinct orientations will be kept separate; a testable corollary is that performance degrades when shapes are genuinely similar and close.
  • If the pseudo-measurement approximation holds, the same recursion should work for other parametric extents, such as rectangles or star-convex shapes, by swapping the shape model and its pseudo-measurement.
  • The claim that Trajectory Set Theory adds no component growth assumes the Gaussian-mixture representation; an extension to non-Gaussian or unknown measurement-rate settings would need to check whether trajectory dimension growth changes complexity.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The manuscript proposes two multi-extended-target tracking filters, the TPHD-E and TCPHD-E, obtained by embedding Trajectory Set Theory (TST) into the extended-target PHD and CPHD recursions, and by combining these with a decoupled elliptical-shape estimation model borrowed from Yang's work. Gaussian mixture implementations (GM-TPHD-E and GM-TCPHD-E) are given, including closed-form recursions, pruning/merging, and state extraction. The methods are evaluated in two scenarios, one simulated and one based on the SIND traffic dataset, using the Gaussian Wasserstein distance and the trajectory metric. The central claim is that these filters provide explicit, accurate ellipse shape estimates and complete, stable trajectory generation, outperforming an RMM-based trajectory PHD filter and a label-based ET-PHD filter.

Significance. If the recursions were correct, the paper would offer a practically useful combination of track-oriented TST filtering with explicit shape estimation for extended targets. A strength of the paper is that it provides full Gaussian-mixture implementations and evaluates on both simulated and real-data scenarios, which goes beyond purely conceptual derivations. However, the published recursions contain several specific mathematical errors that affect the core update equations. The most serious are the PGF-derivative formulas in the GM-TCPHD-E update, which are internally inconsistent with the paper's own Poisson assumptions. Because these formulas feed directly into the posterior cardinality distribution, the printed TCPHD-E recursion does not implement the claimed extended-target CPHD correction. The experimental results therefore cannot be used as evidence for the algorithms as stated, although the errors appear to be correctable.

major comments (4)
  1. [Section IV-B, Eqs. (96)-(98)] The PGF derivative formulas used in the GM-TCPHD-E cardinality update are incorrect. For the clutter PGF G_FA(z)=e^{λ(z−1)}, the n-th derivative at 0 is λ^n e^{−λ}, not n! e^{−λ} as stated in Eq. (96). Similarly, for a target measurement PGF with parameter γ, G_z^{(n)}(0)=γ^n e^{−γ}, not n! e^{−γ} as in Eq. (97). Eq. (98) evaluates G_{k|k−1}^{(|P|)}(υ) as a single term (|P|)! υ^{|P|} P_{ωk}(|P|), whereas the true derivative of the PGF is a sum over all cardinalities m≥|P| of P_{ωk}(m) m!/(m−|P|)! υ^{m−|P|}. These errors propagate through ε_{P,C}, μ_{P,C}, ν_{P,C} and Eq. (22), so the printed cardinality update is not the extended-target CPHD recursion. This is an internal inconsistency, not a modeling choice, and it directly undercuts the central claim of a correct closed-form TCPHD-E recursion.
  2. [Section IV-A, Eq. (55)] The component weight update for the GM-TPHD-E filter appears to have the wrong dependence on the measurement-cell weight. According to the pseudo-likelihood in Eq. (14), the contribution of a measurement cell C is divided by (ϱ(P,C)+ς(P,C)); specifically, L_Z contains φ((P,C)|ξ)/(ϱ(P,C)+ς(P,C)). However, Eq. (55) multiplies the component weight by (ϱ(P,C)+ς(P,C)) instead of dividing by it. The resulting weight w^{(j,P,C)}_{πk} is therefore dimensionally and functionally inconsistent with the continuous-time recursion it is meant to implement. If this is a typographical error, it must be corrected; as printed, the GM-TPHD-E update does not match Eqs. (13)–(17).
  3. [Section IV-B, Eq. (87)] The GM-TCPHD-E cardinality prediction formula contains two errors. The sum over l is written as 'Σ_{l−j}^{∞}' rather than 'Σ_{l=j}^{∞}', and the argument of the posterior cardinality distribution inside the sum is P_{πk−1}(n) rather than P_{πk−1}(l). With these errors, the formula does not sum over the correct index and does not use the prior cardinality distribution correctly. The intended expression is presumably the constant-pS specialization of Eq. (20), which uses P_{πk−1}(l) and a sum from l=j to infinity. As printed, this formula would produce an incorrect predicted cardinality distribution and would invalidate the subsequent update.
  4. [Section IV-A, Eqs. (56)-(58)] The trajectory covariance update in Eqs. (56)–(57) sets the posterior trajectory covariance to a block-diagonal matrix via the construction with I(n^{(j)}_{ωk}), discarding cross-time correlations. The text acknowledges this as a 'slight error' and compares it to a '1-Scan' implementation. However, the paper claims closed-form Bayesian recursions and complete trajectory generation. A block-diagonal covariance after update is an approximation, not an exact Bayesian update for the trajectory state, and the paper does not quantify the resulting error. This assumption should be stated as an approximation in the main derivation and its effect on trajectory accuracy should be assessed, particularly since the experiments report trajectory metrics.
minor comments (4)
  1. [Abstract and Introduction] The phrase 'closed Bayesian recursive' in the abstract should be 'closed-form Bayesian recursions'; the same wording issue appears in the introduction and conclusion.
  2. [Introduction, Section I] The terms 'Semi Positive Definite Matrix' should be 'positive semidefinite matrix'; the standard terminology is used inconsistently.
  3. [Section IV-A, Eq. (65)] The shape matrix S^{(j)}_{ωk} is defined with rotation and diagonal scaling but the associated measurement covariance computation in Eq. (64) would benefit from a brief justification of why the multiplicative noise term S Q_h S^T is inserted in the measurement covariance; the paper currently refers only to [40] without a self-contained explanation.
  4. [Section V, Table IV] The table header uses 'Qe = 1 4 ˙Qe' and similar notation which is ambiguous; the intended meaning should be stated explicitly, e.g., 'Qe = (1/4) times the nominal value Qe'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the derivation combines external published components rather than reducing to its own inputs.

full rationale

After walking the claimed derivation chain, I find no circular step. The TPHD-E and TCPHD-E recursions are assembled from independently published external components: trajectory PHD/CPHD machinery [46], extended-target PHD/CPHD likelihoods [21,22,48,49], and Yang's decoupled ellipse-shape pseudo-measurement model [40]. The paper explicitly attributes the shape update equations (Eqs. 72-85) to [40], and the GM implementations in Propositions 1-4 substitute Gaussian-mixture trajectory densities into these external recursions. None of the load-bearing references are by the present authors, so there is no self-citation chain carrying the argument. The admitted 'slight error' in the block-diagonal trajectory covariance (Eqs. 56-57) is a stated approximation, not a hidden refit of the predicted quantity. The birth model in Scenario 1 is initialized near true positions, but this is an evaluation convenience applied uniformly to all compared filters and is not a fitted parameter renamed as a prediction. The printed PGF derivative formulas in Eqs. 96-98 appear internally suspect as a correctness matter, but a mathematical error is not circularity. Overall, no equation reduces by construction to its own inputs, and the evaluation uses external metrics and baselines.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard trajectory set theory and the decoupled shape model, which are external and cited. The main assumptions are the validity of the linearized shape update in the multi-target setting and the correctness of the published PGF and weight formulas, which are currently in doubt. No new physical or mathematical entities are invented.

free parameters (4)
  • birth weights w_beta = 0.1 per component
    Set by hand in Scenario 1; affects track initiation and cardinality estimates.
  • initial shape state s0 = [0, 45, 35] (Scenario 1), [0, 1.5, 1.5] (Scenario 2)
    Chosen manually; the paper notes it is larger than the true shape, which disadvantages the RMM baseline but also affects the proposed filters.
  • merge thresholds TMr and TMs = 4 and 1
    Pruning and merging thresholds set by hand; no sensitivity analysis is provided.
  • process and measurement noise parameters qr, qtheta, ql, qe = 10, 0.05, 0.1, 10 (Scenario 1)
    Scenario parameters chosen by the authors and applied uniformly to all methods; not fitted to the proposed method specifically.
assumptions (5)
  • domain assumption Trajectory PHD/CPHD recursions from [46] are valid for extended targets when the measurement likelihood depends only on the current target state.
    Used to write Eq. (8)-(14) and to justify multiplying the whole trajectory density by the current-state pseudo-likelihood.
  • domain assumption Extended target measurement model: number of measurements per target is Poisson with constant rate gamma, measurements are uniformly distributed on the target surface, and clutter is Poisson with constant rate lambda.
    Used in Eq. (18) and throughout the measurement likelihood and cardinality updates.
  • domain assumption The decoupled shape model of Yang and Baum [40], with measurement model z = Hr + Sh + e, h ~ N(0, (1/4)I), and the linearized stacked update in Eq. (68)-(85), is an accurate Bayesian update in the multi-target trajectory context.
    Borrowed directly from [40]; the paper does not re-derive or validate its accuracy when embedded in the trajectory PHD/CPHD recursion.
  • ad hoc to paper Kinematic and shape states are Gaussian and remain Gaussian after update, and the block-diagonal trajectory covariance construction in Eq. (56)-(57), which ignores cross-time correlations, is a valid approximation.
    The paper acknowledges the 'slight error' in this construction and cites the 1-Scan implementation in [46], but does not quantify the impact on trajectory quality.
  • ad hoc to paper PGF derivative formulas used in the TCPHD-E update: G^(n)(0) = n! e^{-lambda} for a Poisson PGF and G^(n)(v) = n! v^n P(n).
    These formulas appear in Eq. (96)-(98) and are load-bearing for the cardinality update, but they are mathematically incorrect for a Poisson PGF; the correct derivatives are lambda^n e^{-lambda} and a sum over the cardinality distribution.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The PHD/CPHD filter for Multiple Extended Target Tracking with Trajectory Set Theory and Explicit Shape Estimation." pith.science (2026). https://pith.science/paper/KV4TGDR2

@misc{pith2026250415040,
  author       = {Pith},
  title        = {Pith review of: The PHD/CPHD filter for Multiple Extended Target Tracking with Trajectory Set Theory and Explicit Shape Estimation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KV4TGDR2}},
  note         = {Machine review of arXiv:2504.15040}
}
read the original abstract

In this paper, we propose two methods for tracking multiple extended targets or unresolved group targets with elliptical extent shape. These two methods are deduced from the famous Probability Hypothesis Density (PHD) filter and the Cardinality-PHD (CPHD) filter, respectively. In these two methods, Trajectory Set Theory (TST) is combined to establish the target trajectory estimates. Moreover, by employing a decoupled shape estimation model, the proposed methods can explicitly provide the shape estimation of the target, such as the orientation of the ellipse extension and the length of its two axes. We derived the closed Bayesian recursive of these two methods with stable trajectory generation and accurate extent estimation, resulting in the TPHD-E filter and the TCPHD-E filter. In addition, Gaussian mixture implementations of our methods are provided, which are further referred to as the GM-TPHD-E filter and the GM-TCPHD-E filters. We illustrate the ability of these methods through simulations and experiments with real data. These experiments demonstrate that the two proposed algorithms have advantages over existing algorithms in target shape estimation, as well as in the completeness and accuracy of target trajectory generation.

Figures

Figures reproduced from arXiv: 2504.15040 by the authors.

Figure 1
Figure 1. The Ground Truth (GT) of the Scenario 1. [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. Fig.2. It can be seen that the GM-TPHD-GIW filer, based [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. The TM result for Scenario 1. In addition, to analyze the results more thoroughly, we present the average TM results of the four methods in Fig.3 to evaluate the precision of all methods in generating target trajectories. It can be seen that the GM-TPHD-GIW filter has the highest location cost because of its disadvantage in shape estimation. The GM-LPHD-E filter has a higher switch cost before the 30s, which is the … view at source ↗
Figures from the paper (6 more)
Figure 2
Figure 2. Figure 2: One sample simulation of four methods for Scenario 1. [PITH_FULL_IMAGE:figures/full_fig_p012_2.png]
Figure 4
Figure 4. Figure 4: The GWD result for Scenario 1. More details of Scenario 2 are listed in the Tab.III. Due to the smaller scene size, the following modifications have been made to the filtering parameter settings compared to Scenario 1: • The standard deviation of process noise: q r = 0…
Figure 8
Figure 8. Figure 8: The TM result for Scenario 2. 10 20 30 40 50 60 70 80 90 0 1 2 3 4 5 6 GM-LPHD-E GM-TPHD-GIW GM-TPHD-E GM-TCPHD-E [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]
Figure 9
Figure 9. Figure 9: The estimated cardinality result for Scenario 2 [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]
Figure 6
Figure 6. Figure 6: One example simulation of four methods for Scenario 2. [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: The GWD result for Scenario 2. the lag in trajectory building, peaks appear on its missed cost between 1s and 10s, as well as between 48s and 55s. • The GM-LPHD-E filter has a divergent GWD in Fig.7, due to trajectory switching between Tar. 4 and Tar. 5, this phenomeno…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 52 canonical work pages

  1. [1]

    A tutorial on multiple extended object tracking,

    K. Granstr ¨om and M. Baum, “A tutorial on multiple extended object tracking,” Techrxiv, 2022. [Online]. Available: http://dx.doi.org/10. 36227/techrxiv.19115858.v1

    work page 2022

  2. [2]

    Tracking multiple vehicles using a variational radar model,

    A. Scheel and K. Dietmayer, “Tracking multiple vehicles using a variational radar model,” IEEE Trans. Intell. Transp. Syst. , vol. 20, no. 10, pp. 3721–3736, 2019

    work page 2019

  3. [3]

    Gnn-pmb: A simple but effective online 3d multi-object tracker without bells and whistles,

    J. Liu, L. Bai, Y . Xia, T. Huang, B. Zhu, and Q.-L. Han, “Gnn-pmb: A simple but effective online 3d multi-object tracker without bells and whistles,” IEEE Trans. Intell. Veh., vol. 8, no. 2, pp. 1176–1189, 2023

    work page 2023

  4. [4]

    Autonomous driving at ulm university: A modular, robust, and sensor-independent fusion approach,

    F. Kunz, D. Nuss, J. Wiest, H. Deusch, S. Reuter, F. Gritschneder, A. Scheel, M. St ¨ubler, M. Bach, P. Hatzelmann, C. Wild, and K. Diet- mayer, “Autonomous driving at ulm university: A modular, robust, and sensor-independent fusion approach,” in Proc. 2015 IEEE Intell. Veh. Symp., 2015, pp. 666–673

    work page 2015

  5. [5]

    Scalable detection and tracking of geometric extended objects,

    F. Meyer and J. L. Williams, “Scalable detection and tracking of geometric extended objects,” IEEE Trans. Signal Process. , vol. 69, pp. 6283–6298, 2021

    2021

  6. [6]

    Lxl: Lidar excluded lean 3d object detection with 4d imaging radar and camera fusion,

    W. Xiong, J. Liu, T. Huang, Q.-L. Han, Y . Xia, and B. Zhu, “Lxl: Lidar excluded lean 3d object detection with 4d imaging radar and camera fusion,” IEEE Trans. Intell. Veh., vol. 9, no. 1, pp. 79–92, 2024

    work page 2024

  7. [7]

    Particle filtering for multiple object tracking in dynamic fluorescence microscopy images: Application to microtubule growth analysis,

    I. Smal, K. Draegestein, N. Galjart, W. Niessen, and E. Meijering, “Particle filtering for multiple object tracking in dynamic fluorescence microscopy images: Application to microtubule growth analysis,” IEEE Trans. Med. Imaging, vol. 27, no. 6, pp. 789–804, 2008

    work page 2008

  8. [8]

    Multiple extended target tracking based on glmb filter and gibbs sampler,

    Y . Chen, W. Liu, and X. Wang, “Multiple extended target tracking based on glmb filter and gibbs sampler,” in Proc. 6th Int. Conf. Control, Autom. Inf. Sci., 2017, pp. 26–31

    work page 2017

Show all 56 references
  1. [9]

    Extended object tracking with imm filter for automotive pre- crash safety applications,

    A. Hanumegowda, S. Dewangan, S. Bhupala, F. Gruson, and D. Stein- hauser, “Extended object tracking with imm filter for automotive pre- crash safety applications,” in Proc. 18th European Radar Conf. , 2022, pp. 177–180

    2022

  2. [10]

    Extended object tracking: Introduction, overview and applications,

    K. Granstr ¨om and M. Baum, “Extended object tracking: Introduction, overview and applications,” ArXiv, vol. abs/1604.00970, 2016. [Online]. Available: https://api.semanticscholar.org/CorpusID:14452673

    2016 arXiv

  3. [11]

    On the use of multiple measurement models for extended target tracking,

    K. Granstr ¨om and C. Lundquist, “On the use of multiple measurement models for extended target tracking,” in Proc. 16th Int. Conf. on Inf. Fusion, 2013, pp. 1534–1541

    2013

  4. [12]

    An improved b-spline extended object tracking model using the iterative closest point method,

    K.-M. Dahl ´en, C. Lindberg, M. Yoneda, and T. Ogawa, “An improved b-spline extended object tracking model using the iterative closest point method,” in Proc. 25th Int. Conf. on Inf. Fusion , 2022, pp. 1–8

    2022

  5. [13]

    Tracking of extended object using random triangle model,

    M. Li, J. Lan, and X. R. Li, “Tracking of extended object using random triangle model,” IEEE Trans. Aerosp. Electron. Syst. , vol. 61, no. 2, pp. 2926–2940, 2025

    2025

  6. [14]

    Tracking of rectangular object using key points with regionally concentrated measurements,

    X. Cao, J. Lan, Y . Liu, and B. Tan, “Tracking of rectangular object using key points with regionally concentrated measurements,” IEEE Trans. Intell. Transp. Syst. , vol. 25, no. 6, pp. 5312–5327, 2024

    2024

  7. [15]

    Symmetric star-convex shape tracking with wishart filter,

    H. Alqaderi, F. Govaers, and W. Koch, “Symmetric star-convex shape tracking with wishart filter,” in Proc. 24th Int. Conf. on Inf. Fusion , 2021, pp. 1–8

    2021

  8. [16]

    Extended target tracking using gaussian processes,

    N. Wahlstr ¨om and E. ¨Ozkan, “Extended target tracking using gaussian processes,” IEEE Trans. Signal Process., vol. 63, no. 16, pp. 4165–4178, 2015

    2015

  9. [17]

    Bayesian wishart filter for random shape tracking,

    H. Alqaderi, F. Govaers, and W. Koch, “Bayesian wishart filter for random shape tracking,” IEEE Trans. Aerosp. Electron. Syst. , vol. 58, no. 3, pp. 1941–1952, 2022

    1941

  10. [18]

    Bayesian approach to extended object and cluster tracking using random matrices,

    J. W. Koch, “Bayesian approach to extended object and cluster tracking using random matrices,” IEEE Trans. Aerosp. Electron. Syst. , vol. 44, no. 3, pp. 1042–1059, 2008

    2008

  11. [19]

    Probabilistic data association for tracking extended targets under clutter using random matrices,

    M. Schuster, J. Reuter, and G. Wanielik, “Probabilistic data association for tracking extended targets under clutter using random matrices,” in Proc. 18th Int. Conf. on Inf. Fusion , 2015, pp. 961–968

    2015

  12. [20]

    A pmht approach for extended objects and object groups,

    M. Wieneke and W. Koch, “A pmht approach for extended objects and object groups,” IEEE Trans. Aerosp. Electron. Syst. , vol. 48, no. 3, pp. 2349–2370, 2012

    2012

  13. [21]

    A phd filter for tracking multiple extended targets using random matrices,

    K. Granstr ¨om and U. Orguner, “A phd filter for tracking multiple extended targets using random matrices,” IEEE Trans. Signal Process. , vol. 60, no. 11, pp. 5657–5671, 2012

    2012

  14. [22]

    An extended target cphd filter and a gamma gaussian inverse wishart implementation,

    C. Lundquist, K. Granstr ¨om, and U. Orguner, “An extended target cphd filter and a gamma gaussian inverse wishart implementation,” IEEE J. Sel. Top. Signal Process. , vol. 7, no. 3, pp. 472–483, 2013

    2013

  15. [23]

    Poisson multi-bernoulli mixture conjugate prior for multiple extended target filtering,

    K. Granstr ¨om, M. Fatemi, and L. Svensson, “Poisson multi-bernoulli mixture conjugate prior for multiple extended target filtering,” IEEE Trans. Aerosp. Electron. Syst. , vol. 56, no. 1, pp. 208–225, 2020

    2020

  16. [24]

    Poisson multi-bernoulli approximations for multiple extended object filtering,

    Y . Xia, K. Granstr ¨om, L. Svensson, M. Fatemi, ´A. F. Garc´ıa-Fern´andez, and J. L. Williams, “Poisson multi-bernoulli approximations for multiple extended object filtering,” IEEE Trans. Aerosp. Electron. Syst. , vol. 58, no. 2, pp. 890–906, 2022

    2022

  17. [25]

    Gamma gaussian inverse wishart probability hypothesis density for extended target tracking using x-band marine radar data,

    K. Granstr ¨om, A. Natale, P. Braca, G. Ludeno, and F. Serafino, “Gamma gaussian inverse wishart probability hypothesis density for extended target tracking using x-band marine radar data,” IEEE Trans. Geosci. Remote Sens., vol. 53, no. 12, pp. 6617–6631, 2015

    2015

  18. [26]

    Rfs-based multiple extended target tracking with resolved multipath JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 16 detections in clutter,

    B. Liu, R. Tharmarasa, R. Jassemi, D. Brown, and T. Kirubarajan, “Rfs-based multiple extended target tracking with resolved multipath JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 16 detections in clutter,” IEEE Trans. Intell. Transp. Syst. , vol. 24, no. 10, pp. 1...

    2015

  19. [27]

    Mallick, V

    M. Mallick, V . Krishnamurthy, and B.-N. V o, Bayesian Multiple Target Filtering Using Random Finite Sets . Wiley-IEEE Press, 2012, pp. 75– 126

    2012

  20. [28]

    Data association and track management for the gaussian mixture probability hypothesis density filter,

    K. Panta, D. E. Clark, and B.-N. V o, “Data association and track management for the gaussian mixture probability hypothesis density filter,” IEEE Trans. Aerosp. Electron. Syst. , vol. 45, no. 3, pp. 1003– 1016, 2009

    2009

  21. [29]

    Multiple extended target tracking with labeled random finite sets,

    M. Beard, S. Reuter, K. Granstr ¨om, B.-T. V o, B.-N. V o, and A. Scheel, “Multiple extended target tracking with labeled random finite sets,”IEEE Trans. Signal Process., vol. 64, no. 7, pp. 1638–1653, 2016

    2016

  22. [30]

    Bayesian multiple extended target tracking using labeled random finite sets and splines,

    A. Daniyan, S. Lambotharan, A. Deligiannis, Y . Gong, and W.-H. Chen, “Bayesian multiple extended target tracking using labeled random finite sets and splines,” IEEE Trans. Signal Process. , vol. 66, no. 22, pp. 6076–6091, 2018

    2018

  23. [31]

    Poisson multi-bernoulli mixture trackers: Continuity through random finite sets of trajectories,

    K. Granstr ¨om, L. Svensson, Y . Xia, J. Williams, and ´A. F. Garc´ıa-Fem´andez, “Poisson multi-bernoulli mixture trackers: Continuity through random finite sets of trajectories,” in Proc. 21th Int. Conf. on Inf. Fusion, 2018, pp. 1–5

    2018

  24. [32]

    Spatiotemporal constraints for sets of trajectories with applications to pmbm densities,

    K. Granstr ¨om, L. Svensson, Y . Xia, ´A. F. Garc ´ıa-Fem´andez, and J. Williams, “Spatiotemporal constraints for sets of trajectories with applications to pmbm densities,” in Proc. 23th Int. Conf. on Inf. Fusion , 2020, pp. 1–8

    2020

  25. [33]

    Extended object tracking using sets of trajectories with a phd filter,

    J. Sjudin, M. Marcusson, L. Svensson, and L. Hammarstrand, “Extended object tracking using sets of trajectories with a phd filter,” in Proc. 24th Int. Conf. on Inf. Fusion , 2021, pp. 1–8

    2021

  26. [34]

    The trajectory phd filter for coexisting point and extended target tracking,

    S. Wei, ´A. F. Garc ´ıa-Fern´andez, and W. Yi, “The trajectory phd filter for coexisting point and extended target tracking,” IEEE Trans. Aerosp. Electron. Syst., pp. 1–18, 2025

    2025

  27. [35]

    Trajectory pmb filters for extended object tracking using belief propagation,

    Y . Xia, ´A. F. Garc´ıa-Fern´andez, F. Meyer, J. L. Williams, K. Granstr ¨om, and L. Svensson, “Trajectory pmb filters for extended object tracking using belief propagation,” IEEE Trans. Aerosp. Electron. Syst. , vol. 59, no. 6, pp. 9312–9331, 2023

    2023

  28. [36]

    Poisson multi-bernoulli mixtures for sets of trajectories,

    K. Granstr ¨om, L. Svensson, Y . Xia, J. Williams, and ´A. F. Garc ´ıa- Fern´andez, “Poisson multi-bernoulli mixtures for sets of trajectories,” IEEE Trans. Aerosp. Electron. Syst. , vol. 61, no. 2, pp. 5178–5194, 2025

    2025

  29. [37]

    New prediction for extended targets with random matrices,

    K. Granstr ¨om and U. Orguner, “New prediction for extended targets with random matrices,” IEEE Trans. Aerosp. Electron. Syst. , vol. 50, no. 2, pp. 1577–1589, 2014

    2014

  30. [38]

    Tracking of extended object or target group using random matrix: new model and approach,

    J. Lan and X. R. Li, “Tracking of extended object or target group using random matrix: new model and approach,” IEEE Trans. Aerosp. Electron. Syst., vol. 52, no. 6, pp. 2973–2989, 2016

    2016

  31. [39]

    Extended target tracking with a lidar sensor using random matrices and a virtual measurement model,

    P. Hoher, S. Wirtensohn, T. Baur, J. Reuter, F. Govaers, and W. Koch, “Extended target tracking with a lidar sensor using random matrices and a virtual measurement model,” IEEE Trans. Signal Process., vol. 70, pp. 228–239, 2022

    2022

  32. [40]

    Tracking the orientation and axes lengths of an elliptical extended object,

    S. Yang and M. Baum, “Tracking the orientation and axes lengths of an elliptical extended object,” IEEE Trans. Signal Process., vol. 67, no. 18, pp. 4720–4729, 2019

    2019

  33. [41]

    Random matrix based extended target tracking with orientation: A new model and inference,

    B. Tuncer and E. ¨Ozkan, “Random matrix based extended target tracking with orientation: A new model and inference,” IEEE Trans. Signal Process., vol. 69, pp. 1910–1923, 2021

    1910

  34. [42]

    Linear-time joint probabilistic data association for multiple extended object tracking,

    S. Yang, K. Thormann, and M. Baum, “Linear-time joint probabilistic data association for multiple extended object tracking,” in Proc. 10th Sens. Array Multichannel Signal Proces. Workshop , 2018, pp. 6–10

    2018

  35. [43]

    Marginal association probabilities for multiple extended objects without enumeration of measurement partitions,

    S. Yang, L. M. Wolf, and M. Baum, “Marginal association probabilities for multiple extended objects without enumeration of measurement partitions,” in Proc. 23th Int. Conf. on Inf. Fusion , 2020, pp. 1–8

    2020

  36. [44]

    Network flow labeling for extended target tracking phd filters,

    S. Yang, F. Teich, and M. Baum, “Network flow labeling for extended target tracking phd filters,” IEEE Trans. Ind. Inf. , vol. 15, no. 7, pp. 4164–4171, 2019

    2019

  37. [45]

    Globally-optimal greedy algorithms for tracking a variable number of objects,

    H. Pirsiavash, D. Ramanan, and C. C. Fowlkes, “Globally-optimal greedy algorithms for tracking a variable number of objects,” in Proc. 2011 IEEE Comput. Soc. Conf. Comput. Vision. Pattern. Recognit., 2011, pp. 1201–1208

    2011

  38. [46]

    Trajectory phd and cphd filters,

    ´A. F. Garc ´ıa-Fern´andez and L. Svensson, “Trajectory phd and cphd filters,” IEEE Trans. Signal Process. , vol. 67, no. 22, pp. 5702–5714, 2019

    2019

  39. [47]

    Spatial distribution model for tracking extended objects,

    K. Gilholm and D. Salmond, “Spatial distribution model for tracking extended objects,” in IEE. Proc. Radar Sonar. Navig. , 2005. [Online]. Available: https://api.semanticscholar.org/CorpusID:123307030

    2005

  40. [48]

    Implementation of the giw-phd filter,

    K. Granstr ¨om and U. Orguner, “Implementation of the giw-phd filter,”

  41. [49]

    Extended target tracking with a cardinalized probability hypothesis density filter,

    U. Orguner, C. Lundquist, and K. Granstr ¨om, “Extended target tracking with a cardinalized probability hypothesis density filter,” in Proc. 14th Int. Conf. on Inf. Fusion , 2011, pp. 1–8

    2011

  42. [50]

    Tracking and sensor data fusion,

    W. Koch, “Tracking and sensor data fusion,” 2014. [Online]. Available: https://api.semanticscholar.org/CorpusID:57776404

    2014

  43. [51]

    Trajectory phd filter with unknown detection profile and clutter rate,

    S. Wei, B. Zhang, and W. Yi, “Trajectory phd filter with unknown detection profile and clutter rate,” in Proc. 6th Int. Conf. Signal Image Process., 2021, pp. 1155–1161

    2021

  44. [52]

    Trajectory phd and cphd filters with unknown detection profile,

    ——, “Trajectory phd and cphd filters with unknown detection profile,” IEEE Trans. Veh. Technol., vol. 71, no. 8, pp. 8042–8058, 2022

    2022

  45. [53]

    Metrics for performance evaluation of elliptic extended object tracking methods,

    S. Yang, M. Baum, and K. Granstr ¨om, “Metrics for performance evaluation of elliptic extended object tracking methods,” in IEEE Int. Conf. Multisensor Fusion Integr. Intell. Syst. , 2016, pp. 523–528

    2016

  46. [54]

    A metric on the space of finite sets of trajectories for evaluation of multi-target tracking algorithms,

    ´A. F. Garc´ıa-Fern´andez, A. S. Rahmathullah, and L. Svensson, “A metric on the space of finite sets of trajectories for evaluation of multi-target tracking algorithms,” IEEE Trans. Signal Process. , vol. 68, pp. 3917– 3928, 2020

    2020

  47. [55]

    Sind: A drone dataset at signalized intersection in china,

    Y . Xu, W. Shao, J. Li, K. Yang, W. Wang, H. Huang, C. Lv, and H. Wang, “Sind: A drone dataset at signalized intersection in china,” in Proc. 25th Conf. Intell. Transport Syst. , 2022, pp. 2471–2478

    2022

  48. [2012]

    Available: https://api.semanticscholar.org/CorpusID: 264670906

    [Online]. Available: https://api.semanticscholar.org/CorpusID: 264670906

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.