Pith. sign in

REVIEW 3 major objections 5 minor 39 references

Development of a Robotic System for Automatic Wheel Removal and Fitting

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read A two-stage vision pipeline detects a wheel hub and tracks it across lighting and distance.

desk verdict A competent but purely qualitative demo of standard OpenCV algorithms on a single wheel; the vision claims are plausible but unmeasured, and the title promises more than the paper delivers. read the letter →

arxiv 1908.09009 v1 pith:PJM5NWSR submitted 2019-08-19 cs.CV cs.ROeess.IV

classification cs.CVcs.ROeess.IV
keywords CircularHoughTransformCamshifttrackingwheelhubdetectionrobotictyrechangingOpenCVreal-timeobjectcolourhistogramback-projectionvision-guidedrobotics
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 is a feasibility study for the vision half of a robotic wheel-changing system. Its central claim is that a two-stage computer vision pipeline — Circular Hough Transform (CHT) to detect the wheel hub, then Continuously Adaptive Mean Shift (Camshift) colour tracking to follow it — can keep a tracking window on the hub of a sample vehicle wheel under different room lighting conditions and at distances from about 1 m to 3 m from a laptop webcam. The application that motivates the work is replacing manual tyre changing, which causes injuries. If the claim holds, the detected hub centre and radius give a robot arm the reference it needs to approach, remove, and refit a wheel.

What carries the argument

The load-bearing machinery is the pairing of two algorithms. CHT in its Hough-gradient form casts votes in a two-dimensional accumulator to find circle centres and radii, and its success depends on tuning parameters such as the minimum distance between centres and the maximum radius. Camshift extends mean-shift tracking by back-projecting the target's colour histogram onto each frame and computing zeroth-, first-, and second-order image moments; the zeroth moment sets the new search-window size, so the window can grow or shrink as the target moves closer or farther. That self-resizing behaviour is what lets the same tracker follow the hub at 1 m and the larger wheel silhouette at 2 to 3 m.

What would settle it

Record a video of the wheel moving from 3 m to 1 m under the same lighting, label the wheel hub centre in every frame, and compare the Camshift window's tracked centre to those labels; if the centre error exceeds roughly the hub radius for more than a small fraction of frames, the claim that the tracker effectively follows the hub fails.

Watch

Extended reading notes

Core claim

The paper reports that CHT, with manually tuned parameters (accumulator resolution, Canny threshold, minimum distance between centres, and radius limits), detected the wheel hub as a circle of radius 34 pixels and the full tyre as a circle of radius 215 pixels in webcam images. Using the detected hub as the region of interest, Camshift's continuously adaptive colour histogram then kept the tracking window on the hub at about 1 m under lightly dim, moderately lit, and well-lit room lighting. When the wheel was moved to about 2 m and 3 m, the tracking window resized itself and readjusted to track the entire wheel rather than just the hub, consistent with Camshift's adaptive window-size mechanism. The paper's stated conclusion is that these algorithms provide software solutions that can be deployed with a robotic mechanical arm to make tyre changing faster, safer, and more efficient.

Load-bearing premise

The paper treats success on selected still frames of one sample wheel, with parameters tuned until they worked, as evidence that the tracker is accurate and real-time enough for a robot to remove and fit wheels in general.

Editorial extensions

If this is right

  • A robot controller can use the detected hub centre and radius as a target reference for aligning a gripper or wrench with the wheel.
  • Because Camshift updates its colour distribution every frame, the tracker can absorb slow changes in lighting without reinitialization.
  • Restricting processing to the hub region of interest keeps computation light enough for real-time tracking.
  • For a fixed camera and a known wheel, tuning CHT parameters once may suffice; wrong settings produce false circles, as the paper demonstrates.

Reading between the lines

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

  • A quantitative version of this test would matter: reporting per-frame centre error against labelled ground truth, and frames per second, would turn the qualitative claim into a threshold that robot controllers can trust.
  • Because Camshift is colour-based, the method should transfer well to wheels with a distinctive hub colour, but would likely struggle if hub and background share a colour histogram; adding an edge or depth cue could fix that.
  • The same detection-tracking chain could be tested on lug-nut positions as the wheel rotates, which would let the robot orient the wheel before fitting; the paper does not address rotation.
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

3 major / 5 minor

Summary. The paper proposes a vision component for a robotic wheel-removal and fitting system. It applies the Circular Hough Transform (CHT) in OpenCV to detect a circular wheel hub (and the outer tyre wheel) in a single webcam image, and then uses the Camshift algorithm to track the hub region across video frames under three qualitatively described lighting conditions and at distances of about 1 m, 2 m, and 3 m. The reported results are qualitative: still frames from Figures 15–17 are used to argue that the tracker follows the wheel hub at 1 m and adapts its window to the whole wheel at larger distances. The conclusion generalizes from this single-wheel, manually tuned demonstration to the feasibility of fully automated robotic wheel changing.

Significance. If the central claim were quantitatively established, the paper would document a minimal vision capability for a wheel-changing robot: CHT-based detection of a wheel hub as an ROI and Camshift-based tracking of that ROI in webcam video across varying lighting and distance. The paper has the merit of applying established, reproducible algorithms (OpenCV CHT and Camshift) and of clearly disclosing the CHT parameter adjustments in Tables 1–3, so there is no hidden derivation or circularity in the sense of fitting a model and then presenting the fit as validation. However, the significance is currently limited because the headline claims of 'accurate' and 'real-time' tracking are supported only by inspection of a few still frames; no quantitative tracking error, frame rate, false-positive statistics, or independent test images are reported. The contribution is therefore a qualitative demonstration rather than a validated system component.

major comments (3)
  1. [§IV.C, Figures 15–17] The central claim that Camshift 'could effectively track the wheel hub' at about 1 m and 'continued to track' at 2 m and 3 m is not supported by quantitative evidence. No tracking error is reported in any form: there is no center-coordinate error, no intersection-over-union with a ground-truth hub box, no failure count, and no frame-rate measurement, so the terms 'effective' and 'real time' are unfalsifiable as stated. In addition, the paper's own descriptions indicate a target-identity change: at 2 m and 3 m the window 'resizes and readjusts itself to track the entire wheel,' meaning the tracker initialized on the hub drifts to the whole wheel. This is scale drift rather than hub tracking unless the authors show that the hub center remains within the window, which they do not. For a robotic manipulator, the vision output must localize the hub or wheel center with known accuracy, so this missing metric is load-bearing.
  2. [§IV.B, Tables 1–3] The CHT detection stage is demonstrated on a single sample wheel image with parameters manually adjusted until the desired circle is found (Table 2 for the hub, Table 3 for the tyre). No held-out images, no false-positive rate, and no variation of lighting, pose, wheel type, or camera distance are presented for the detection stage. The claim that the system can 'accurately detect and classify specific objects of interest' therefore rests on one hand-tuned example, and the transfer of these fixed parameters to an industrial robotic setting is an assumption rather than a demonstrated result.
  3. [§IV.C and Conclusion] The paper overgeneralizes from a single wheel sample, a fixed webcam, and manually tuned CHT parameters to the feasibility of 'fully automated robotic systems' for wheel changing. There is no integration with a manipulator, no wheel-removal or fitting experiment, and no test of the vision output as control feedback. The conclusion should be restricted to the demonstrated capability—qualitative tracking of one wheel under tested conditions—or the paper needs additional experiments showing end-to-end performance.
minor comments (5)
  1. [§II.C and §III] Equation numbering is duplicated: the kernel definition in §II.E is numbered (1), but equation (1) in §II.C is already the circle equation. Please renumber sequentially.
  2. [§IV.A, Figures 11 and 14] The center coordinate of the detected hub is given as [378 292] in Figure 11 but as (292,378) in the captions of Figures 13 and 14; the order should be made consistent and the axis convention stated.
  3. [References] References [34] and [37] are the same paper (Wang and Li), and both are cited for related claims; this duplicate should be merged or distinguished.
  4. [§IV.C] The lighting conditions are described only as 'lightly dim room,' 'moderately lit room,' and 'well-lit room.' Reporting quantitative illumination values or at least a reproducible setup would improve the paper's reproducibility.
  5. [Various] There are several grammatical and typographical issues, for example 'tyre' and 'tire' are used interchangeably in places, and some figure captions are incomplete sentences ('Performance of Camshift Algorithm at about 1m from the Laptop webcam under different lighting condition'). A careful proofread is needed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper applies standard CHT and Camshift algorithms with disclosed parameter tuning, and its load-bearing claims do not reduce to fitted inputs or self-citations.

full rationale

The paper is an empirical application of well-established, externally documented algorithms: Circular Hough Transform for circle detection and Camshift for tracking. The CHT stage uses manually adjusted parameters (Tables 1-3) until the desired circle is found; this is disclosed parameter tuning on the same sample image, not fitting a parameter to a subset of data and presenting a closely related quantity as an independent prediction. The Camshift stage initializes its search window from the CHT-derived ROI and reports qualitative still frames at different distances and lighting conditions. That evaluation lacks quantitative error metrics and ground truth, so the central claim is weakly verified and potentially overgeneralized, but a verification weakness is not circularity. No load-bearing step rests on a self-citation chain: the cited works, including Bradski's Camshift paper and the Kalman-Camshift comparison in [37], are prior external sources used for context and agreement, not uniqueness theorems or assumptions that make the result true by construction. The paper's 'predictions' are not equivalent to its inputs: the CHT output is not the same object as the tracking claim, and the tracking window self-adjustment is an algorithmic behavior, not a derived result that presupposes the conclusion. Accordingly, no specific circular reduction can be exhibited, and the appropriate finding is no significant circularity.

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

The numerical content rests on five hand-tuned CHT parameters and on standard CHT/Camshift implementations from the cited OpenCV literature. No new theoretical quantities are introduced and no invented entities are needed. The domain assumptions about color stability and single-wheel representativeness are the least supported parts.

free parameters (5)
  • CHT inverse ratio of accumulator resolution to image resolution (Dp) = 0.1 for hub/tyre; 0.8 in the false circle test
    Chosen by hand per test image to obtain the desired circle; no sensitivity analysis is provided.
  • CHT minimum distance between detected circle centers (minDist) = 18 pixels for final detections; 150 pixels in the false circle test
    Adjusted manually to suppress false detections on the sample; the paper states the 150-pixel value produced 5 false circles.
  • Canny edge detector higher threshold (param1) = 50 for final detections
    Hand-chosen threshold affecting which edges enter the Hough accumulator.
  • CHT accumulator threshold (param2) = 33 for final detections; 20 in the false circle test
    Hand-tuned to reject false circle centers while keeping the true center.
  • CHT maximum circle radius = 50 pixels for hub; 0 (unset) for tyre
    Set by hand to capture the hub and prevent false large circles; the generality of this bound is not tested.
assumptions (5)
  • standard math CHT circle equation (x-xc)^2 + (y-yc)^2 = r^2 and the OpenCV Hough gradient implementation find the relevant circle parameters.
    Used in Section II.C-D as the detection mechanism; treated as standard background from [10]-[18].
  • standard math Camshift moment and mean-shift equations (Eq. 3-8) define the target center and window resizing.
    Used in Section III as the tracking mechanism; standard algorithm from [30]-[36].
  • domain assumption The wheel hub's color histogram is stable enough to be a tracking signature across dim, moderate, and well-lit rooms.
    Invoked implicitly in Section IV.C, where the ROI color histogram is used to track, but no color constancy or illumination analysis is provided.
  • ad hoc to paper A single sample wheel and manually set CHT parameters represent the operating range of industrial wheel-changing robots.
    The Conclusion generalizes to fully automated systems from one sample wheel, distances of 1-3 m, and one webcam, without validation on other wheels or environments.
  • domain assumption The wheel remains within the camera field of view and the dominant color of the hub persists during tracking.
    Section IV.C states explicitly that the target must remain in view for tracking to hold; this is an environmental constraint not enforced by the system.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Development of a Robotic System for Automatic Wheel Removal and Fitting." pith.science (2026). https://pith.science/paper/PJM5NWSR

@misc{pith2026190809009,
  author       = {Pith},
  title        = {Pith review of: Development of a Robotic System for Automatic Wheel Removal and Fitting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PJM5NWSR}},
  note         = {Machine review of arXiv:1908.09009}
}
read the original abstract

This paper discusses the image processing and computer vision algorithms for real time detection and tracking of a sample wheel of a vehicle. During the manual tyre changing process, spinal and other muscular injuries are common and even more serious injuries have been recorded when occasionally, tyres fail (burst) during this process. It, therefore, follows that the introduction of a robotic system to take over this process would be a welcome development. This work discusses various useful applicable algorithms, Circular Hough Transform (CHT) as well as Continuously adaptive mean shift (Camshift) and provides some of the software solutions which can be deployed with a robotic mechanical arm to make the task of tyre changing faster, safer and more efficient. Image acquisition and software to accurately detect and classify specific objects of interest were implemented successfully, outcomes were discussed and areas for further studies suggested.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages

  1. [1]

    Automation and Artificial Intelligence - Brookings Institution,

    R. M. a. J. W. M Muro, "Automation and Artificial Intelligence - Brookings Institution," 2019

  2. [2]

    Automatic Tire Changing of Large Mining Vehicles with Industrial Robots,

    H. Staab et al. , "Automatic Tire Changing of Large Mining Vehicles with Industrial Robots," in Proceedings of ISR 2016: 47st International Symposium on Robotics, 2016: VDE, pp. 1-7

  3. [3]

    A Robotic Parallel Platform for Automated Tire Changing of Large Mining Vehicles,

    H. Staab and J. Newkirk, "A Robotic Parallel Platform for Automated Tire Changing of Large Mining Vehicles," in ISR/Robotik 2014; 41st International Symposium on Robotics, 2014: VDE, pp. 1-7

  4. [4]

    Macro - micro manipulation with visual tracking and its application to wheel assembly,

    C. Cho, S. Kang, M. Kim, and J. -B. Song, "Macro - micro manipulation with visual tracking and its application to wheel assembly," International Journal of Control, Automation, and Systems, vol. 3, no. 3, pp. 461-468, 2005

  5. [5]

    Two -stage control approach of a robot manipulator for conveyor tracking application,

    T. Park, B . Lee, and I. H. Suh, "Two -stage control approach of a robot manipulator for conveyor tracking application," in Proceedings of IECO N'94-20th Annual Conference of IEEE Industrial Electronics , 1994, vol. 2: IEEE, pp. 691-696

  6. [6]

    Control of a ro botic manipulator to grasp a moving target using vision,

    N. Houshangi, "Control of a ro botic manipulator to grasp a moving target using vision," in Proceedings., IEEE International Conference on Robotics and Automation, 1990: IEEE, pp. 604-609

  7. [7]

    An adaptive robotic tracking system using optical flow,

    R. C. Luo, R. Mullen, and D. E. Wessell, "An adaptive robotic tracking system using optical flow," in Proceedings. 1988 IEEE International Conference on Robotics and Automation, 1988: IEEE, pp. 568-573

  8. [8]

    Corke, Robotics, vision and control: fundamental algorithms in MATLAB® second, completely revised

    P. Corke, Robotics, vision and control: fundamental algorithms in MATLAB® second, completely revised . Springer, 2017

Show all 39 references
  1. [9]

    Introduction to digital image and video processing,

    A. C. Bovik, "Introduction to digital image and video processing," in Handbook of image and video processing: Elsevier, 2005, pp. 3-I

  2. [10]

    Find circles using circular Hough transform

    "Find circles using circular Hough transform." Mathworks.https://uk.mathworks.com/help/images/re f/imfindcircles.html (accessed 27/06/2019, 2019)

  3. [11]

    Method and means for recognizing complex patterns,

    P. V. Hough, "Method and means for recognizing complex patterns," ed: Google Patents, 1962

  4. [12]

    ªUse of the Hough Transform to Detect Lines and Curves in Pictures, º Comm,

    R. Duda and P. Hart, "ªUse of the Hough Transform to Detect Lines and Curves in Pictures, º Comm," ed: ACM, 1972

  5. [13]

    Comparative study of Hough transform methods for circle f inding,

    H. Yuen, J. Princen, J. Illingworth, and J. Kittler, "Comparative study of Hough transform methods for circle f inding," Image and vision computing, vol. 8, no. 1, pp. 71-77, 1990

  6. [14]

    Circular object detection using a modified Hough transform,

    M. Smereka and I. Dulęba, "Circular object detection using a modified Hough transform," International Journal of Applied Mathematics and Computer Science, vol. 18, no. 1, pp. 85-91, 2008. a. Lightly dim room b. Moderately lit room c. Well-lit room 8

  7. [15]

    O'Reilly Media, Inc

    G. Bradski and A. Kaehler, Learning OpenCV: Computer vision with th e OpenCV library. " O'Reilly Media, Inc.", 2008

  8. [16]

    Brahmbhatt, Practical OpenCV

    S. Brahmbhatt, Practical OpenCV. Apress, 2013

  9. [17]

    Laganière, OpenCV Computer Vision Application Programming Cookbook Second Edition

    R. Laganière, OpenCV Computer Vision Application Programming Cookbook Second Edition . Packt Publishing Ltd, 2014

  10. [18]

    Hough Circle Transform

    "Hough Circle Transform." https://docs.opencv.org/2. 4/doc/tutorials/imgproc/imgtrans/hough_circle/hough _circle.html (accessed 15/07/2019, 2019)

  11. [19]

    Review and evaluation of well -known methods for moving object detection and tracking in videos,

    B. Karasulu, "Review and evaluation of well -known methods for moving object detection and tracking in videos," Journal of aeronautics and space technologies, vol. 4, no. 4, pp. 11-22, 2010

  12. [20]

    Structural similarity-based object tracking in video sequences,

    A. Loza, L. Mihaylova, N. Canagarajah, and D. Bull, "Structural similarity-based object tracking in video sequences," in 2006 9th International Conference on Information Fusion, 2006: IEEE, pp. 1-6

  13. [21]

    Robust techniques for background subtraction in urban traffic video,

    S. C. Sen -Ching and C. Kamath, "Robust techniques for background subtraction in urban traffic video," in Visual Communications and Image Processing 2004 , 2004, vol. 5308: International Society for Optics and Photonics, pp. 881-892

  14. [22]

    Review a nd evaluation of commonly-implemented background subtraction algorithms,

    Y. Benezeth, P.-M. Jodoin, B. Emile, H. Laurent, and C. Rosenberger, "Review a nd evaluation of commonly-implemented background subtraction algorithms," in 2008 19th International Conference on Pattern Recognition, 2008: IEEE, pp. 1-4

  15. [23]

    Object tracking: A survey,

    A. Yilmaz, O. Javed, and M. Shah, "Object tracking: A survey," Acm computing surveys (CSUR), vol. 38, no. 4, p. 13, 2006

  16. [24]

    Tracking and classifying moving objects from video,

    Q. Zhou and J. K. Aggarwal, "Tracking and classifying moving objects from video," in Proceedings of IEEE Workshop on Performance Evaluation of Tracking and Surveillance , 2001, vol. 12: Hawaii, USA

  17. [25]

    Mean shift: A robust approach toward feature space analysis,

    D. Comaniciu and P. Meer, "Mean shift: A robust approach toward feature space analysis," IEEE Transactions on Pattern Analysis & Machine Intelligence, no. 5, pp. 603-619, 2002

  18. [26]

    Mean shift, mode seeking, and clustering,

    Y. Cheng, "Mean shift, mode seeking, and clustering," IEEE transactions on pattern analysis and machine intel ligence, vol. 17, no. 8, pp. 790 -799, 1995

  19. [27]

    The estimation of the gradient of a density function, with applications in pattern recognition,

    K. Fukunaga and L. Hostetler, "The estimation of the gradient of a density function, with applications in pattern recognition," IEEE Transactions on information theory, vol. 21, no. 1, pp. 32-40, 1975

  20. [28]

    B. W. Silverman, Density Estimation for Statistics and Data Analysis. CRC Press, 1986

  21. [29]

    Conceptual clustering in knowledge organization,

    Y. Cheng and K.-S. Fu, "Conceptual clustering in knowledge organization," IEEE Transactions on Pattern Analysis and Machine Intelligence, no. 5, pp. 592-598, 1985

  22. [30]

    Object tracking using improved CAMShift algorithm combined with motion segmentation,

    E. Ema mi and M. Fathy, "Object tracking using improved CAMShift algorithm combined with motion segmentation," in 2011 7th Iranian Conference on Machine Vision and Image Processing , 2011: IEEE, pp. 1-4

  23. [31]

    Computer vision face tracking for use in a perceptual user interface,

    G. R. Bradski, "Computer vision face tracking for use in a perceptual user interface," 1998

  24. [32]

    Fast and robust Camshift tracking,

    E. David, B. Erich, K. Daniel, and S. Anselm, "Fast and robust Camshift tracking," IEEE Transaction, vol. 8, pp. 1-8, 2010

  25. [33]

    Object tracking using camshift algorithm and multiple quantized feature spaces,

    J. G. Allen, R. Y. Xu, and J. S. Jin, "Object tracking using camshift algorithm and multiple quantized feature spaces," in Proceedings of the Pan -Sydney area workshop on Visual information p rocessing, 2004: Australian Computer Society, Inc., pp. 3-7

  26. [35]

    Camshift object tracking algorithm implementation on DM6437 EVM,

    A. Varfolomieiev, O. Antonyuk, and O. Lysenko, "Camshift object tracking algorithm implementation on DM6437 EVM," in The 4th European DSP in Education and research conference, 2010: Citeseer

  27. [36]

    Open source computer vision library reference manual,

    C. Intel, "Open source computer vision library reference manual," Intel Corporation Microprocessor Research Labs, 2001

  28. [37]

    The study of MovingTarget tracking based on Kalman-CamShift in the video,

    X. Wang and X. Li, "The study of MovingTarget tracking based on Kalman-CamShift in the video," in The 2nd International Conference on Information Science and Engineering, 2010: IEEE, pp. 1-4

  29. [38]

    CAMSHIFT improvement on multi-hue and multi-object tracking,

    P. Hidayatullah and H. Konik, "CAMSHIFT improvement on multi-hue and multi-object tracking," in Proceedings of the 2011 International Conference on Electrical Engineering and Informatics , 2011: IEEE, pp. 1-6

  30. [39]

    Real -time tracking method for moving target based on an improved Camshift algorithm,

    Y.-B. Li, X. -l. Shen, and S. -s. Bei, "Real -time tracking method for moving target based on an improved Camshift algorithm," in 2011 International Conference on Mechatronic Science, Electric Engineering and Computer (MEC) , 2011: IEEE, pp. 978-981

  31. [40]

    Object tracking with Camshift – OpenCV 3.4 with python 3 Tutorial 30,

    S. Canu, "Object tracking with Camshift – OpenCV 3.4 with python 3 Tutorial 30," ed, 2018

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.