Pith. sign in

REVIEW 4 major objections 5 minor 38 references

Event-based multi-view photogrammetry for high-dynamic, high-velocity target measurement

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This paper claims that an event-based multi-view photogrammetric system, built on monotonicity-based leading-edge extraction and reprojection-error matching, measures a high-velocity fragment's initial velocity with a 4.47% deviation from…

desk verdict A useful event-camera photogrammetry pipeline for high-velocity fragments, with a coherent but under-validated 4.47% accuracy claim that needs more experiment reporting before I'd trust the number. read the letter →

arxiv 2506.00578 v1 pith:55BQEKM6 submitted 2025-05-31 cs.CV

Taihang Lei , Banglei Guan , Minzu Liang , Xiangyu Li , Jianbing Liu , Jing Tao , Yang Shang , Qifeng Yu This is my paper
classification cs.CV
keywords eventcamerahigh-velocityfragmentmulti-viewphotogrammetryleading-edgefeatureextractionreprojectionerrorvelocitydecaymodelmotionmeasurementneuromorphicvision
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

The paper tries to show that an array of event cameras can replace high-speed cameras and contact sensors for measuring very fast, small projectiles. Because event cameras respond to brightness changes with microsecond latency and high dynamic range, they can observe a fragment that would overexpose or blur ordinary cameras. The proposed pipeline isolates the fragment's leading edge from confusing tailing events, matches events to a 3D trajectory with reprojection error, and fits a fragment velocity-decay model to the matched timestamps. In a light-gas-gun test it reports a fitted initial velocity of 385.6 m/s, within 4.47% of an electromagnetic speedometer reading and 1.80% of a gas-gun theory prediction.

What carries the argument

The load-bearing mechanism is the leading-edge feature extraction driven by monotonicity in the event spatiotemporal distribution. Events near the trajectory are projected into a distance-versus-time histogram; because the fragment's distance from the scattering origin increases monotonically, an event farther from the origin than the current reference becomes the new reference, and the resulting reference set is the leading edge, free of tailing events. This cleaned set feeds a reprojection-error association step: the 3D trajectory is obtained by line-based intersection, search points along it are projected with the pinhole model, and each event is bound to the nearest projection within a 2-pixel reception threshold, producing timestamped 3D points. The measurement then rests on the fragment velocity-decay model v = v0/(1+k v0 t) and its displacement integral, which are fitted to the timestamped points to yield the initial velocity.

What would settle it

Place two speedometer stations or a calibrated high-speed camera at known positions along the 1 m trajectory and compare the fitted velocity curve at each station; if the fitted curve matches only one station or the displacement-fit residuals show systematic curvature, the decay model or the leading-edge timing is incorrect.

Watch

Extended reading notes

Core claim

The paper's central claim is that asynchronous event streams from two calibrated event cameras can yield continuous, high-time-resolution trajectory data for a high-velocity fragment, and fitting a theoretical velocity-decay curve to those data recovers the fragment's initial velocity. The discovery is a measurement paradigm: instead of relying on point correspondences and stereo intersection, the method first extracts leading-edge feature events by exploiting the monotonic increase of distance from the scattering origin, then associates each event with a 3D point along the fitted trajectory by minimizing reprojection error, and finally fits the displacement-time integral of v = v0/(1+k v0 t) to estimate v0. On a light gas gun firing an 8 mm, roughly 5 g fragment, the method yields v0 = 385.6 m/s compared with 369.1 m/s from an electromagnetic speedometer (4.47% deviation) and 378.8 m/s from gas-gun theory (1.80% deviation), while providing 213 usable photogrammetric points versus 6 points for traditional event-based stereo intersection.

Load-bearing premise

The central assumption is that the fragment's velocity decay over the observed flight region follows v = v0/(1+k v0 t) with a constant decay coefficient k, so extrapolating the fitted curve back to launch time yields the same initial velocity the electromagnetic speedometer reports.

Editorial extensions

If this is right

  • The method supplies continuous trajectory data for a 1 m flight region, so velocity decay can be observed along the path rather than only at a single light screen.
  • Reprojection-error association yields 213 usable timestamped 3D points, compared with 6 points for a traditional 20 microsecond-window stereo intersection, which stabilizes the velocity fit.
  • The event-camera pipeline operates at 1 microsecond event resolution, equivalent to roughly 146,000 fps, so a short observation window still constrains the motion model.
  • Because the system avoids damped witness plates and uses cameras priced around a tenth of high-speed cameras, it offers a lower-cost route to high-dynamic mechanical testing.

Reading between the lines

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

  • A co-located reference, such as a second speedometer station or a calibrated high-speed camera along the same trajectory, would separate the optical method's own error from any mismatch between the speedometer's measurement location and the fit's extrapolation to t=0; the reported 4.47% deviation is likely an upper bound on the optical error rather than a clean accuracy measure.
  • The monotonicity-based leading-edge extraction generalizes beyond fragments: any target whose distance from a fixed origin grows monotonically could use the same rule to suppress tailing events, including shock fronts, projectiles, or moving vehicles.
  • Fitting the same timestamped data with a two-parameter decay model or checking residuals against a power-law drag law would reveal whether the one-parameter decay model is adequate over the short observation window.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes an event-based multi-view photogrammetric pipeline for measuring the motion of high-velocity fragments. It consists of three stages: leading-edge event extraction based on spatiotemporal monotonicity (Section 3.2), event-to-trajectory association via reprojection error (Section 3.3), and velocity-decay-model fitting to recover the initial velocity (Section 3.4). The method is tested in a light gas gun experiment with two event cameras observing a 1 m dispersion region; the fitted initial velocity is 385.6 m/s, reported as a 4.47% deviation from an electromagnetic speedometer reading of 369.1 m/s and a 1.80% deviation from a theoretical prediction of 378.8 m/s. The paper claims continuous observation, high time resolution (equivalent ~146,000 fps), and 213 matched data points versus 6 for traditional stereo intersection.

Significance. If the accuracy claim were properly validated, the work would be a useful engineering contribution. It demonstrates a low-cost, high-dynamic-range alternative to high-speed cameras for ballistic velocity measurement, with a coherent algorithmic pipeline and a real experiment against an independent sensor. The leading-edge extraction addresses the tailing effect that complicates event-based motion measurement, and the reprojection-error association is a sensible way to use more of the event stream. The paper is clearly organized and describes the algorithm in enough detail to be reproduced. However, the evidence base is a single shot with an incompletely specified reference measurement, so the magnitude of the claimed advance is not yet established.

major comments (4)
  1. [Section 4.3, Table 1] The headline accuracy claim (4.47% deviation) is not a clean measure of the photogrammetric method because the electromagnetic speedometer's measurement location, averaging baseline, and measurement uncertainty are never stated. Under the paper's own decay model, Eq. (12) with k = cxρs/(2M) ≈ 0.006 m^-1 (using M=5 g, ρ=1.2 kg/m^3, s≈5×10^-5 m^2, cx≈1), the observed gap of 16.5 m/s between 385.6 and 369.1 m/s corresponds to roughly 7.3 m of additional flight distance. Since the cameras observe only a 1 m region and the speedometer could plausibly be mounted downstream, the deviation may be dominated by physical velocity decay rather than by measurement error. The authors must report the speedometer's position relative to the muzzle, its timing baseline and method, and its calibration uncertainty, and should compare against the fitted velocity at the same location.
  2. [Table 1] The relative deviations in Table 1 use different references: 4.47% is 16.5/369.1 (relative to the speedometer), while 1.80% is 6.80/378.8 (relative to the theoretical prediction). Since the abstract quotes 4.47% as 'compared to the electromagnetic speedometer,' the theory row should use the same reference, which would give 2.6% relative to the speedometer rather than 1.80%. The direction of the deviation (Ours minus reference) should also be stated explicitly to avoid ambiguity.
  3. [Section 4.3 and Section 3.4] The validation is based on a single shot: there are no repeated firings, no error bars on the fitted v0 or k, and no uncertainty propagation from camera calibration or event timing. The conclusion that the method demonstrates 'good accuracy and stability' is therefore unsupported as written. At minimum, the authors should provide the fitting residuals, confidence intervals for the fitted parameters, and the speedometer's specified accuracy; ideally, the experiment should be repeated several times to establish repeatability.
  4. [Section 3.4] The initial velocity v0 is obtained by extrapolating the decay model v = v0/(1 + k v0 t) to t=0 from data collected over a roughly 1 m trajectory. This extrapolation assumes that the drag law (constant k, no Mach-number or tumbling effects) holds over the whole flight, but no independent check of the model is given. Because the speedometer comparison is the only external validation, a mismatch could be absorbed by an incorrect decay law; the authors should verify the model shape (for example, by comparing local velocities computed from short time differences with the model prediction) or quantify the sensitivity of v0 to the assumed k.
minor comments (5)
  1. [Section 3.2, Algorithm 1] Algorithm 1 does not initialize the loop variable c before the while condition; it should be initialized to 2 (since the first event is the reference) to match the description in the text.
  2. [Section 3.3, Eq. (6)] Equation (6) contains a typographical error: 'Ωi and is the recorded value' should read 'Ωi is the recorded value.'
  3. [Section 4.3, Table 1] The label 'Electro metric' in Table 1 should be 'Electromagnetic speedometer,' and the entry for the speedometer should include its measurement uncertainty if available.
  4. [Section 4.3] The claim of an 'equivalent frame rate around 146,000 fps' is not derived; if it comes from the 1 µs event timestamp resolution, the authors should clarify the reasoning, because asynchronous events do not directly translate to a frame rate.
  5. [Section 4.3] The comparison with 'traditional event-based stereo intersection' uses a 20 µs time window and a 2-pixel epipolar constraint; the authors should explain why these parameters are representative and, ideally, compare the velocity estimate obtained from the 6-point subset to quantify the benefit of the 213-point set.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported initial velocity is a fitted model parameter validated against an independent electromagnetic speedometer.

full rationale

The paper's central output is the fitted initial velocity v0 = 385.6 m/s obtained by fitting the standard ballistic decay model v = v0/(1 + kv0t) (Eq. 8) to photogrammetric position-time data. The accuracy claim is a direct comparison of this fitted value to an independent electromagnetic speedometer reading (369.1 m/s), and the two values are produced by completely separate measurement chains. No parameter of the event-camera pipeline is calibrated to the speedometer output, and the decay-model parameters (v0, k) are free parameters of a least-squares fit rather than quantities defined in terms of the comparison target. The trajectory-association step (Section 3.3) reconstructs the 3D trajectory by intersecting 2D line fits of leading-edge events and then assigns timestamps via reprojection error; although the same events are used both to define the trajectory and to provide the timestamps, this does not force v0 to any preset value and the final fit is constrained only by the assumed physical decay model. The paper's self-citations (e.g., Refs. [12], [14], [15]) are contextual and are not load-bearing for the derivation of the velocity result. The validation weakness identified in the skeptic note (unspecified speedometer location and uncertainty) concerns experimental completeness, not circularity. Therefore no circular step exists and the score is 0.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The central claim depends on the monotonicity assumption for leading-edge extraction, the ballistic decay model for velocity fitting, the pinhole camera model with calibrated parameters, and several hand-set thresholds (reception threshold 2 px, search step 0.01 mm). The initial velocity v0, decay coefficient k, and integration constant C are fitted to the event data; their fitted values (except v0) are not reported. No new physical entities are introduced.

free parameters (5)
  • v0 (initial velocity) = 385.6 m/s
    The initial velocity is the primary output of the fit in Section 3.4 (Eq. 10), using the Trust Region algorithm on 213 associated points.
  • k (velocity decay coefficient) = not reported
    The decay coefficient k in Eq. (8/9) appears to be fitted along with v0, but its fitted value is not reported; if fixed from literature, that is an unstated assumption.
  • C (integration constant) = not reported
    The constant C in Eq. (10) is a free fit parameter that absorbs the unknown initial displacement offset.
  • Reception threshold Omega = 2 px
    The reception threshold is set to 2 pixels based on the maximum calibration reprojection error (1.67 pixels) in Section 3.3; this is a hand-chosen parameter.
  • Search step size along trajectory = 0.01 mm
    The step size for searching along the 3D trajectory is set to 0.01 mm; it is chosen without sensitivity analysis.
assumptions (3)
  • domain assumption The distance between the target and the scattering origin increases monotonically over time.
    Used in Section 3.2 and Eq. (3) to extract leading-edge features; fails if the projected pixel distance from the origin decreases (e.g., target passing near a camera or strong noise).
  • domain assumption The fragment velocity follows the decay law v = v0 / (1 + k v0 t) with constant k (Eq. 8).
    Adopted from ballistics ref [29]; the basis for the displacement fit in Section 3.4 and for extrapolating v0; assumes constant drag coefficient and no tumbling or Mach-dependent drag.
  • standard math The event camera is accurately modeled by the pinhole projection (Eqs. 4-5) with parameters from Zhang's calibration.
    Standard camera model; also the reprojection-error association in Section 3.3 assumes calibration accuracy within 1.67 pixels.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Event-based multi-view photogrammetry for high-dynamic, high-velocity target measurement." pith.science (2026). https://pith.science/paper/55BQEKM6

@misc{pith2026250600578,
  author       = {Pith},
  title        = {Pith review of: Event-based multi-view photogrammetry for high-dynamic, high-velocity target measurement},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/55BQEKM6}},
  note         = {Machine review of arXiv:2506.00578}
}
read the original abstract

The characterization of mechanical properties for high-dynamic, high-velocity target motion is essential in industries. It provides crucial data for validating weapon systems and precision manufacturing processes etc. However, existing measurement methods face challenges such as limited dynamic range, discontinuous observations, and high costs. This paper presents a new approach leveraging an event-based multi-view photogrammetric system, which aims to address the aforementioned challenges. First, the monotonicity in the spatiotemporal distribution of events is leveraged to extract the target's leading-edge features, eliminating the tailing effect that complicates motion measurements. Then, reprojection error is used to associate events with the target's trajectory, providing more data than traditional intersection methods. Finally, a target velocity decay model is employed to fit the data, enabling accurate motion measurements via ours multi-view data joint computation. In a light gas gun fragment test, the proposed method showed a measurement deviation of 4.47% compared to the electromagnetic speedometer.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

38 extracted references · 37 canonical work pages

  1. [1]

    Y. Gao, Z. Huang, G. Li, and Y. Xing, A novel stiffness prediction method with constructed microscopic displacement field for periodic beam-like

    work page

  2. [2]

    C. W. Gardiner, Handbook of Stochastic Methods for Physics,

    work page

  3. [3]

    Idiart, M.I., Fleck, N.A.: Size effects in the torsion of thin metal

    work page

  4. [4]

    F. Lu, X. Li, Z. Tian, B. Jiang, and X. Chen, Weapon damage and assessment, Science Press, 2021

    work page 2021

  5. [5]

    Z. Wang, F. Lu, and X. Duan, Design and evaluation of missile tests, Science Press, 2022

    work page 2022

  6. [6]

    J. Wu, M. Qiao, Z. Yan, Y. Jia, Q. Luo, W. Sun, and M. Ren, Overview of testing technology for warhead fragments characterization, Journal of Sichuan Ordnance, pages 105-109, 2019

    work page 2019

  7. [7]

    J. Zhu, Y. Zheng, W.n Li, Y. Yang, X. Wang, X. Qiao, and R. Li, Axial distribution of fragments from the dynamic explosion fragmentation of metal shells, International Journal of Impact Engineering, pages 140-146, 2019

    work page 2019

  8. [8]

    Y. Li, Y. Li, and Y. Wen, Radial distribution of fragment velocity of asymmetrically initiated warhead, International Journal of Impact Engineering, pages 39-47, 2017

    work page 2017

Show all 38 references
  1. [9]

    X. He, J. Liang, X. Wang, X. Wang, and S. Yu, Overview of conventional warhead dynamic explosion power research, Protective Engineering, pages 1-9, 2022

    2022

  2. [10]

    Mahowald, VLSI analogs of neuronal visual processing: a synthesis of form and function, PhD thesis, California Institute of Technology, 1992

    M. Mahowald, VLSI analogs of neuronal visual processing: a synthesis of form and function, PhD thesis, California Institute of Technology, 1992

    1992

  3. [11]

    Y. Zhou, Z. Yan, Y. Yang, Z. Wang, P. Lu, P. Yuan, and B. He, Bioinspired sensors and applications in intelligent robots: a review, Robotic Intelligence and Automation, pages 215-228, 2024

    2024

  4. [12]

    Gallego, T

    G. Gallego, T. Delbruck, G. Orchard, C. Bartolozzi, B. Taba, A. Censi, S. Leutenegger, A. Davison, J. Conradt, K. Daniilidis, and D. Scaramuzza, Event-based vision: a survey, IEEE transactions on pattern analysis and machine intelligence, pages 154-180, 2022

    2022

  5. [13]

    D. D. Cho and T. Lee, A review of bioinspired vision sensors and their applications, Sensors and Materials, volume 27, number 6, pages 447--463, 2015

    2015

  6. [14]

    Finateu, A

    T. Finateu, A. Niwa, D. Matolin, K. Tsuchimoto, A. Mascheroni, E. Reynaud, P. Mostafalu, F. Brady, L. Chotard, F. LeGoff, H. Takahashi, H. Wakabayashi, Y. Oike, and C. Posch, 5.10 a 1280×720 back-illuminated stacked temporal contrast event-based vision sensor with 4.86um pixel...

    2020

  7. [15]

    T. Lei, B. Guan, M. Liang, Z. Liu, J. Liu, Y. Shang, and Q. Yu, Motion measurements of explosive shock waves based on an event camera, Optics Express, pages 15390-15409, 2024

    2024

  8. [16]

    H. Zhou, Z. Shi, H. Dong, S. Peng, Y. Chang, and L. Yan, JSTR: joint spatio-temporal reasoning for event-based moving object detection, 2024 IEEE International Conference on Robotics and Automation (ICRA), 2024

    2024

  9. [17]

    Z. Liu, B. Guan, Y. Shang, Y. Bian, P. Sun, and Q. Yu, Stereo event-based, 6-dOF pose tracking for uncooperative spacecraft, IEEE Trans. Geosci. Remote Sens., pages 1-13, 2025

    2025

  10. [18]

    Z. Liu, B. Guan, Y. Shang, Q. Yu, and L. Kneip, Line-based 6-doF object pose estimation and tracking with an event camera, IEEE Trans. Image Process, pages 4765-4780, 2024

    2024

  11. [19]

    J. Ren, H. Jiang, J. Ji, and J. Su, Research summary of test technologies for scattering distribution of warhead fragment field, Journal of Sichuan Ordnance, pages 122-131, 2024

    2024

  12. [20]

    Jiang, X

    J. Jiang, X. Xu, W. Ren, X. Yu, and M. Yao, Overview of fragmentation power field of warhead under dynamic explosion, Acta Armamentarii, pages 1-8, 2023

    2023

  13. [21]

    Z. Yan, J. Wu, X. He, X. Cui, H. Lv, and W. Sun, Review on the development of power field testing technology of high-energy warhead, Journal of Physics: Conference Series, pages 012159, 2023

    2023

  14. [22]

    J. Zhow, D. Kong, C. Xu, Y. Yu, and Y. Tang, Array-type integrated measurement system for fragment velocity and scatter peculiarity, Electronic Measurement Technology, pages 133-140, 2022

    2022

  15. [23]

    J. Chen, J. Su, J. Liu, J. Zhang, X. Liu, and S. Li, Fragment the space coordinate testing system based on active lighting double camera intersection method, Science Technology and Engineering, pages 173-176,187, 2017

    2017

  16. [24]

    J. Ni, J. Fan, and Y. An, Design and experiment of an optical screen by using line structured light laser, Journal of Xi'an Technological University, pages 103-108,135, 2011

    2011

  17. [25]

    L. Qian, Y. Chen, X. Shen, Y. Zhu, and L. Bai, Design and application of the storage-type speed test device for fragment, Journal of Ordnance Equipment Engineering, pages 114-117, 2016

    2016

  18. [26]

    J. Ni, H. Lu, and H. Tian, Research on a method of measuring the impact location dispersion of double barrel cannon based on the seven-light-screen array, Acta Armamentarii, pages 398-405, 2013

    2013

  19. [27]

    G. Liu, Y. Xu, and Q. Li, Research on velocity test of fragment group of gradient scattering preformed Fragment Warhead, Journal of Physics: Conference Series, pages 082021, 2024

    2024

  20. [28]

    Zhang, H

    G. Zhang, H. Zhang, Z. Shen, D. Kong, C. Ning, F. Shang, and X. Zhang, A novel detection method for warhead fragment targets in optical images under dynamic strong interference environments, Defence Technology, pages 252-270, 2025

    2025

  21. [29]

    Zhou and D

    J. Zhou and D. Kong, Research on a measurement method for spatio-temporal distribution of fragments based on a high-speed camera network, Measurement Science and Technology, pages 105406, 2023

    2023

  22. [30]

    Erkai, K

    W. Erkai, K. Nico, P. Robin, M. Hans-Gerd, and H. Stefan, Four-view split-image fragment tracking in hypervelocity impact experiments, International Journal of Impact Engineering, pages 103405, 2020

    2020

  23. [31]

    Yu and Y

    Q. Yu and Y. Shang, Videometrics: principles and researches, Science Press, 2009

    2009

  24. [32]

    Y. Ma, J. Hu, Z. Feng, J. Yang, K. Zhi, and H. Wang, Experimental study on velocity attenuation law of fragment simulated projectile, Journal of Sichuan Ordnance, issue 2, pages 54-61, 2024

    2024

  25. [33]

    D. Tan, D. Wen, and Z. Zhang, Experimental studies on flight performances of spherical fragment, Acta Aerodynamica Sinica, issue 2, pages 263-266, 2008

    2008

  26. [34]

    Zhang, A flexible new technique for camera calibration, Pattern Analysis and Machine Intelligence, IEEE Transactions on, pages 1330-1334, 2000

    Z. Zhang, A flexible new technique for camera calibration, Pattern Analysis and Machine Intelligence, IEEE Transactions on, pages 1330-1334, 2000

    2000

  27. [35]

    H. Su, J. Zhao, and S. Peng, Analysis of the influencing factors of the initial velocity of gas gun projectile, Journal of Projectiles, Rockets, Missiles and Guidance, pages 105-108,113, 2018

    2018

  28. [36]

    J. Liu, D. Guo, Y. Ren, Y. Xu, and X. Li, Air resistance coefficient and flight speed characteristics of spherical fragments, Transactions of Beijing Institute of Technology, issue 7, pages 676-684, 2023

    2023

  29. [37]

    , " * write output.state after.block = add.period write newline

    ENTRY address author booktitle chapter doi edition editor eid howpublished institution journal key month note number organization pages publisher school series title type url volume year label INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION in...

  30. [38]

    write newline

    " write newline "" before.all 'output.state := FUNCTION if.digit duplicate "0" = swap duplicate "1" = swap duplicate "2" = swap duplicate "3" = swap duplicate "4" = swap duplicate "5" = swap duplicate "6" = swap duplicate "7" = swap duplicate "8" = swap "9" = or or or or or or...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.