All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
Thanks to the authors for their efforts to improve the work. I believe this version has addressed the concerns of the reviewers. It is ready for acceptance. Congrats!
[# PeerJ Staff Note - this decision was reviewed and approved by Jyotismita Chaki, a PeerJ Section Editor covering this Section #]
I am okay with the provided responses and action taken.
Authors addressed related concerns!
Authors addressed related concerns!
NA
**PeerJ Staff Note:** Please ensure that all review, editorial, and staff comments are addressed in a response letter and that any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate.
The title is too general, needs to be specific, and map the problem and the methodology
The abstract should contain: Background, objective, problem, methodology, brief findings, brief conclusion, and recommendations
Keywords are ok
The introduction is ok
LR is ok
The paper structure needs improvement
References are not updated
The objective is not clear enough
The problem is clear
The Methodology is too long
The Contribution is not clear
The result and discussion must be separated, and analysis is needed
Conclusion is ok
ok
Add technical clarity by specifying sensor specs, scan times, and algorithm settings.
Improve result interpretation by comparing segmentation errors to industry standards and discussing pattern-specific biases.
Address limitations like axis-aligned cuts, sparse point cloud coverage, and single-segment Gage R&R focus.
The experimental design of the study is robust and well-structured, effectively combining simulation-based validation with real-world testing. The use of modeling clay as a surrogate material is practical for minimizing waste and cost. However, the authors should consider validating the clay’s physical properties (e.g., density and compressibility) against those of actual meat to strengthen the reliability of the results. The decision to test multiple cutting patterns and apply robust statistical methods such as ANOVA and Gage R&R adds rigor to the evaluation. However, focusing Gage R&R analysis on only the first segment may underestimate variability across the entire portioning process, evaluating multiple or randomly selected segments would provide a more comprehensive assessment. The absence of a benchmark comparison, such as manual cutting or an existing algorithm, limits the ability to contextualize the model’s performance. Lastly, the limited sample size in real meat trials, particularly for fish, should be addressed to enhance the study’s generalizability. Overall, the design is strong for a proof-of-concept but would benefit from expanded biological validation and additional controls.
The study's findings are largely valid and well-supported by the experimental data and statistical analyses. The algorithm's ability to segment objects into equal-weight portions was demonstrated through repeated trials with modeling clay, followed by validation using real meat samples (chicken and fish). The use of Mean Absolute Error (MAE), One-Way ANOVA, and Gage R&R provides solid quantitative evidence for the system’s accuracy, repeatability, and reproducibility, demonstrating the thoroughness of the study.
However, the validity is somewhat constrained by certain limitations. The algorithm assumes uniform density and uses only top-surface point cloud data, which may not fully capture natural irregularities in meat texture and thickness. This could potentially lead to inaccuracies in the segmentation process. Additionally, real meat testing involves a limited number of samples, and relying on the first segment for Gage R&R analysis may not fully reflect system variability, potentially underestimating the system's performance.
In summary, the findings are valid within the controlled experimental scope; however, broader claims, especially those related to industrial deployment, would require more extensive biological testing and real-time field validation, a process of testing the system in real-world conditions to ensure its practicality and reliability.
Detailed comments
Introduction (Lines 44–137)
L64–66: “Although initial automation costs are high, long term benefits include…” i would suggest authors to briefly quantifying cost savings or adding a citation supporting the 30% reduction.
L85–89: The explanation of deep learning and 3D point cloud analysis is important, I would suggest authors to consider expanding slightly on how PointNet is adapted for this application.
L119–126: Clearly state that your algorithm assumes uniform density and primarily processes top-surface data (acknowledged later, but worth highlighting early).
Materials and Methods (Lines 139–376)
L139–144: Clarifying the scope, can you also clarify that the contribution lies in algorithmic development, not a full industrial system.
It would also be a good idea to clarify early in the manuscript that real meat testing is only used for final validation (see Line number 395), as readers may expect biological material from the start.
L145–153: Use of Gocator 2140D-3R-R-01-T is appropriate and justified. However could you please include sensor specifications such as resolution (X, Y, Z), field of view, scanning speed. This allows assessment of its suitability for real-time industrial use.
L150: “laser that projected the cutting line”: could authors clarify that the laser was for visualizing cuts, not performing cuts. State if the laser wavelength/power could damage real meat or if it's suitable for commercial settings.
L170–175: Add a brief rationale for using a jig and servo-cylinder. How does this ensure consistent placement, and is this precision critical to cutting accuracy?
Reword the caption for figure 3 as the current captions don’t fully explain how these parts work together.
L194–198: Good wiring description but consider adding a block diagram (sensor → encoder → controller → PC) to clarify data flow. HALCON 23.11 Student Edition—note if this version is limited in functionality (e.g., max image size, export limits).
L204–208: if authors can provide the computational hardware specification which would be very beneficial for the reproducibility. Consider stating the processing time per scan/cut cycle, which is not currently addressed anywhere.
L223–228: Patch-based interpolation method is mentioned, but please specify which algorithm or module in HALCON was used. Readers might benefit from knowing if it was nearest neighbor, bilinear, or B-spline-based.
L254–258: You mention a smoothing technique and multi-resolution approach. Please give some quantitative settings used (e.g., number of smoothing iterations, mesh resolution) or explain how you balanced accuracy vs. computation time.
L260–261: Vertical/horizontal logic is reasonable, but the algorithm assumes cuts must be axis-aligned. Please discuss this as a limitation—freeform or curved cuts could better match organic geometries.
L275–278: “Volume of each vertical segment i = nh[i] * VI” — this assumes all horizontal cuts within a vertical segment are equal. Consider noting if this caused any bias in final results.
L336–339: Final output not only includes cut positions, but these could be used to control real cutters. It would be a good idea if authors can nention if coordinates are stored in a format usable by automated arms or machinery (e.g., CSV, DXF, G-code).
Results and Discussion
L421–433: Volume Segmentation Errors. Good explanation of how sparse data causes under-segmentation. However, the impact of ±0.5% deviation needs a bit more explanation such as if this is acceptable in meat portioning industries? Are there published tolerance ranges (e.g., poultry 5%)?
L456–489: As authors have mentioned that 3.54% error is highest. Please compare this to industrial standards (e.g., typical supermarket product tolerance). Is this acceptable?
L498: To my surprise, an effect size of 0.852 is quite large. Please explain how this was derived (e.g., from pilot study or from literature).
L516–526: You conclude that some patterns perform worse. Please explain why. Is it due to shape asymmetry, scanning angle, or mesh resolution?
L543–547: The choice of “first segment only” is practical, but this may overlook total-system variance. Recommend mentioning this as a design trade-off.
L603–604: "Regardless of small variations in sample placement" — please quantify “small.” Was positioning variation ±5 mm? ±10 degrees?
L627–637: Nice comparison between fish and chicken results. Suggest emphasizing that fish morphology lends itself more to accurate segmentation due to lower surface variation.
Conclusions (L645–662)
L658–659: “...integrating deep learning models for improved decision making...” → Consider giving a specific example (e.g., learning from variable thickness/density data).
L656–658: Mention possible use in other food sectors (e.g., dairy, bakery) if applicable, to show scalability.
Figures and Tables
Some figure legends could be slightly more descriptive (e.g., explain colors in segmentation diagrams).
Table formatting for Table 5: align decimals and ensure consistency across rows.
The manuscript is generally well written, and the English is clear. The literature is cited properly. However, some of the content currently included in the Results and Discussion section should be moved to the Materials and Methods section. It is also unclear how many samples were used. Based on the text, I inferred that 25 samples were included, which is too few to support the conclusion that the system is dependable and reliable.
I recommend a major revision to improve the manuscript structure and an increase in sample size before it can be considered for a second review.
The experimental design appears solid; however, the sample size is a major concern.
sample size is a very low number to draw valid conclusions.
Very long introduction.
Lines 45–55 could be deleted as they contain well-known information. A more focused and relevant literature review should be retained.
Research objectives need revision.
They should specify which industry is being targeted—poultry, fish, or red meat? The literature should also be revised accordingly. After reading the manuscript, it appears the focus is on poultry and fish; therefore, all references to other industries should be removed to avoid confusion.
Title needs revision.
It should clearly reflect the manuscript content, possibly by including terms like "white meat," "poultry," or "fish."
Lines 135–137
This content is unnecessary and should be deleted.
Lines 140–143
These lines do not belong in the Materials and Methods section. They should be moved to the end of the Introduction as the stated objectives of the study.
Line 165
The citation for the software being used is missing and should be added.
Section 2.1
Requires clarification on how the machine vision system is connected to the base. It should also explain how the base receives instructions to coordinate its positioning with the laser cut—specifically, how it measures movement distance and ensures alignment with the laser line profile sensor.
Lines 393–399
It is unclear whether real chicken meat was used. If clay modeling was used instead, the title and corresponding sections must clearly state this. The actual density of meat may differ from that of clay, which impacts the findings. Additionally, the number of repetitions performed should be specified here.
Line 417
The statement “The following section presents the results obtained from the evaluation methods described in Section 2.3, along with a discussion of the findings” is redundant. Readers already understand the purpose of the Results and Discussion section.
Results and Discussion section
Some methodological details are presented here that should instead be placed in the Materials and Methods section. For example:
Lines 467–470: MAE should be defined under M&M.
Section 3.2.1: The calculation and description of %MAE should first be introduced in M&M.
Lines 465–512: This entire section is better suited for M&M rather than Results and Discussion.
Line 494: Verb tense should be reviewed and corrected.
Lines 528–559: Similarly, this section belongs under M&M.
Line 616: This content also belongs in M&M. Furthermore, the number of samples tested is not clearly stated. From the ANOVA table, I inferred that 25 samples were used, which is a very low number to draw reliable conclusions. This brings the reliability of the experiment into question.
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.