Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The submitted article addresses the relevant issue of automating the detection of the whitefly (Bemisia tabaci) on yellow sticky traps using computer vision methods, a topic of significant importance for integrated pest management systems in agriculture. The research objective is to improve insect detection accuracy through image preprocessing with quaternion Gabor filters, subsequent angular optimization, and the application of the YOLOv8 detector. The scientific novelty lies in the combination of a quaternion representation of color images (encoding RGB channels as the pure imaginary part of a quaternion) with Gabor filters to enhance the textural and morphological features of the whitefly, as well as in the adaptive selection of filter orientation angles, which, according to the authors, improves detection metrics compared to traditional methods.

However, the article contains several significant shortcomings and inconsistencies:

- There is ambiguity regarding the dataset used. The authors mention collecting 284 images within the CIPM PeMaTo-Europe project and cite an open dataset [62]. It is not specified whether this specific dataset or proprietary data was employed. With a size of 284 images containing 5,807 instances of whitefly, the dataset is relatively small for training deep neural networks, posing a risk of overfitting despite the use of a pre-trained YOLOv8 model.

- The comparison with other methods in Table 1 is incorrect. It is not indicated on which data the metrics for YOLOv5x, YOLOv7, Faster R-CNN, and other models were obtained. If these figures are taken from the literature without unifying the dataset and experimental conditions, the comparison is invalid. Moreover, the proposed approach (0.95 mAP@0.5) underperforms compared to YOLOv7 (0.968), a point that requires explanation.

- Unsubstantiated quantitative claims are made. The conclusions state a reduction in false positives by 18% and an improvement in mAP by 12.7%. However, it is not specified which baseline variant (YOLOv8 without preprocessing?) this comparison is made against. Furthermore, Table 1 presents a value of 0.957 for YOLOv8s, which is higher than the proposed 0.95, contradicting the claimed improvement.

- The appendix should be removed, as it contains cumbersome, non-informative tables (A1–A3) without any accompanying analysis.

- Figures 2 through 7 should be reduced by half, as they provide little informational value.

- The reference to Figure 4 is missing in the text, and the subsequent numbering of figures in the in-text citations appears to be incorrect.

- The formulas require verification. The parameters in equation (1) are not defined. Equation (8) does not correspond to the classical uncertainty relation for Gabor filters. All formulas should be terminated with appropriate punctuation.

- Tautological phrasing should be revised, for example: "The temporal response of the two-dimensional real Gabor filter may be articulated as the temporal response of the genuine two-dimensional Gabor filter may be articulated…" and "…necessary for the acquisition of large amounts of data for processing in the image acquisition process…"

In conclusion, while the article addresses a promising topic and proposes an original combination of methods, it requires substantial revision prior to publication due to an inadequately developed experimental basis, incorrect comparisons, poor quality of illustrative material, and careless presentation of mathematical expressions.

Author Response

hank you very much for taking the time to review this manuscript. We sincerely appreciate your valuable time, careful attention, and thoughtful consideration of our work. We fully recognize the effort involved in evaluating a research manuscript, and we are truly grateful for your willingness to contribute your knowledge and expertise to this process.

Your review, comments, and suggestions are highly appreciated, as they are essential for improving the quality, clarity, and academic rigor of this study. We humbly value the opportunity to benefit from your observations, and we are grateful for any recommendations that may help us strengthen the manuscript.

Comments 1: There is ambiguity regarding the dataset used. The authors mention collecting 284 images within the CIPM PeMaTo-Europe project and cite an open dataset [62]. It is not specified whether this specific dataset or proprietary data was employed. With a size of 284 images containing 5,807 instances of whitefly, the dataset is relatively small for training deep neural networks, posing a risk of overfitting despite the use of a pre-trained YOLOv8 model.

Response 1: For the first observation, the following paragraph is drafted to unify the sources of the dataset used: "together with Wageningen University, whose participation in the image acquisition process and insect labeling was key to obtaining the batch of images." “[updated text in the manuscript if necessary]”

Comments 2: The comparison with other methods in Table 1 is incorrect. It is not indicated on which data the metrics for YOLOv5x, YOLOv7, Faster R-CNN, and other models were obtained. If these figures are taken from the literature without unifying the dataset and experimental conditions, the comparison is invalid. Moreover, the proposed approach (0.95 mAP@0.5) underperforms compared to YOLOv7 (0.968), a point that requires explanation.

Response 2: change a “this method achieves mAP@0.5 = 0.950, while some approaches report slightly higher values (e.g., YOLOv7: 0.968), however, this proposal incorporates preprocessing with Gabor filters that: reduces false positives in the presence of visual noise (variable lighting, non-target particles), improves the detection of occluded specimens and small nymphs (<0.3 mm), maintains the morphological feature structure of the pest (wings, body) even in adverse conditions, therefore, although the mAP metric is higher, our system is more robust in real field environments, which is critical for precision agriculture applications.”

Comments 3. The appendix should be removed, as it contains cumbersome, non-informative tables (A1–A3) without any accompanying analysis. Response 3: Response: Regarding the appendices, they are limited to the diagnostic table of only the first image of the batch, from which a report will be generated where the system is intended to be an insect counting assistant, classify them, and provide a report on organic pest control techniques. Comments 4: The reference to Figure 4 is missing in the text, and the subsequent numbering of figures in the in-text citations appears to be incorrect. Response 4: for the case of figure 4, the recommendations that were made to us have been addressed Comments 5: The formulas require verification. The parameters in equation (1) are not defined. Equation (8) does not correspond to the classical uncertainty relation for Gabor filters. All formulas should be terminated with appropriate punctuation. Response 5: Response: Mathematical Correction: Equation 8 was updated to the classical uncertainty relation (1/16pi^2), and all parameters for Equation 1 were explicitly defined Comments 6: Tautological phrasing should be revised, for example: "The temporal response of the two-dimensional real Gabor filter may be articulated as the temporal response of the genuine two-dimensional Gabor filter may be articulated…" and "…necessary for the acquisition of large amounts of data for processing in the image acquisition process…" Response 6: In the case where tautology appears in the text, it was rewritten in such a way that the redundancy was eliminated. Comments 7: In conclusion, while the article addresses a promising topic and proposes an original combination of methods, it requires substantial revision prior to publication due to an inadequately developed experimental basis, incorrect comparisons, poor quality of illustrative material, and careless presentation of mathematical expressions. Response 7: Thank you very much for your thoughtful and constructive comments on our manuscript. We greatly appreciate the time and effort you have dedicated to reviewing our work.We have carefully considered all of your recommendations. In response to your observations, we have thoroughly revised the manuscript: we strengthened the experimental basis, corrected the comparative analysis (focusing on our own ablation study rather than direct literature comparisons), improved the quality of all figures and illustrative material, and carefully reviewed and corrected the presentation of mathematical expressions. We believe these changes have substantially improved the clarity and rigor of the paper. We are deeply grateful for your valuable feedback, which has helped us enhance the quality of our work. Thank you again for your respectful and insightful review.   Sincerely, The Authors

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Regarding the article titled: Optimization of Gabor Filters Based on Quaternions for Image Preprocessing in the Automated Detection of Bemisia tabaci in Yellow Traps, I have the following comments.
1. The comparisons should have a description of the variables below them. Without them, misinterpretation is possible.
2. Figure 1 has a font size that is too small. It is currently difficult to read.
3. in 3.1. Image acquisition it is not clear what images are in question. Please indicate the type and manufacturer of the camera and its spectral bands.
4. The Python library YOLO is indicated, but the version of the program that was used is not described.
5. The Discussion section should be expanded. Make a more detailed discussion of your work. If there are similar studies (although I have not solved it) or methodologies, it is good to compare them with your work.
6. Table 2 could be moved to the Discussion section.
The article includes References that are both topical and relevant to the topic.

Author Response

Thank you very much for taking the time to review this manuscript. We sincerely appreciate your valuable time, careful attention, and thoughtful consideration of our work. We fully recognize the effort involved in evaluating a research manuscript, and we are truly grateful for your willingness to contribute your knowledge and expertise to this process.

Your review, comments, and suggestions are highly appreciated, as they are essential for improving the quality, clarity, and academic rigor of this study. We humbly value the opportunity to benefit from your observations, and we are grateful for any recommendations that may help us strengthen the manuscript.

Thank you again for your generosity, professionalism, and dedication to the advancement of scientific research.

Comments and Suggestions for Authors Regarding the article titled: Optimization of Gabor Filters Based on Quaternions for Image Preprocessing in the Automated Detection of Bemisia tabaci in Yellow Traps, I have the following comments. Comment 1. The comparisons should have a description of the variables below them. Without them, misinterpretation is possible. Response 1: "Thank you for this suggestion. We agree that defining the variables is essential for clarity. We have added a descriptive note below Table 2 and Table 3 (Page 17) explaining the metrics mAP@0.5, Precision, Recall, and F1-score. Comment 2. Figure 1 has a font size that is too small. It is currently difficult to read. Response: "We appreciate the observation. The font size in Figure 1 has been increased to ensure all labels are clearly readable at 100% zoom (Page 6)."are clearly readable at 100% zoom (Page 6)." Comment 3. in 3.1. Image acquisition it is not clear what images are in question. Please indicate the type and manufacturer of the camera and its spectral bands. Response 3: "Following your recommendation, we have updated Section 3.1 (Page 7, Lines 150-155) to include the specific hardware details: a Canon EOS 600D camera with an RGB CMOS sensor, operating in the visible spectral bands (Red, Green, and Blue)." Comment 4. The Python library YOLO is indicated, but the version of the program that was used is not described.   Response 4: "Agree. We have clarified in Section 3.5 (Page 8) that the implementation utilized the Ultralytics YOLOv8 library running on Python 3.10." Comment 5. The Discussion section should be expanded. Make a more detailed discussion of your work. If there are similar studies (although I have not solved it) or methodologies, it is good to compare them with your work. Response 5: "We have expanded the Discussion section (Section 5, Page 16) to include a more detailed comparison with other methodologies. Specifically, we now discuss how the quaternionic approach maintains color correlations that are lost in traditional grayscale processing.   The article includes References that are both topical and relevant to the topic. We greatly appreciate the time and effort you have dedicated to reviewing our work. We have carefully considered all of your recommendations. In response to your observations, we have thoroughly revised the manuscript: we strengthened the experimental basis, corrected the comparative analysis (focusing on our own ablation study rather than direct literature comparisons), improved the quality of all figures and illustrative material, and carefully reviewed and corrected the presentation of mathematical expressions. We believe these changes have substantially improved the clarity and rigor of the paper. We are deeply grateful for your valuable feedback, which has helped us enhance the quality of our work. Thank you again for your respectful and insightful review.   Sincerely, The Authors

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Is relevant and the proposed idea is potentially interesting. In its current form the manuscript suffers from serious methodological, reporting and writing issues that prevent a reliable assessment of its scientific contribution. 
A major revision is required with particular attention to reproducibility, consistency of results, stronger experimental validation and substantial language editing.

The authors should clarify the exact scientific novelty of the study and explicitly distinguish the contribution of quaternion Gabor preprocessing from that of a conventional image enhancement plus YOLO v8  through a clear ablation study.

The manuscript should provide a fully reproducible experimental protocol and i suggest to include the exact train-validation-test split, final preprocessing parameters, fixed YOLO v8 hyperparameters, augmentation settings, inference thresholds.

The authors should carefully revise the manuscript to resolve inconsistencies in the reported performance metrics especially the discrepancy between the mAP values reported in different sections of the paper.

The Results section should be strengthened with more rigorous quantitative evaluation under identical experimental conditions. And include baseline comparisons, repeated runs, standard deviations and a clearer error analysis of false positives and false negatives.

The manuscript requires substantial language editing and structural revision. Several sections are currently written in a speculative or incomplete manner and some figure captions and subsections are not presented in a publication ready form.

The authors should reduce the excessive theoretical background and devote more space to the practical implementation details and experimental validation of the proposed method.

The limitations imposed by the relatively small and imbalanced dataset should be discussed more explicitly and the claims regarding robustness and generalization should be moderated accordingly or supported by additional external validation.

Several strong claims should either be supported by direct experimental evidence and statistical justification or be rewritten in a more cautious and scientifically grounded manner.

The conclusion section should be revised so that its claims remain strictly aligned with the results actually demonstrated in the study, without extending beyond the presented evidence.

Author Response

Comments and Suggestions for Authors Comment 1: Is relevant and the proposed idea is potentially interesting. In its current form the manuscript suffers from serious methodological, reporting and writing issues that prevent a reliable assessment of its scientific contribution. A major revision is required with particular attention to reproducibility, consistency of results, stronger experimental validation and substantial language editing. Response1: Thank you for your valuable feedback. We have performed a major revision focusing on methodological clarity, internal consistency, and linguistic quality. We have restructured the experimental section to ensure reproducibility and validated our results with more rigorous statistical evidence. Comment 2: The authors should clarify the exact scientific novelty of the study and explicitly distinguish the contribution of quaternion Gabor preprocessing from that of a conventional image enhancement plus YOLO v8  through a clear ablation study. Response 2: We have clarified that the novelty lies in the use of a hypercomplex (quaternionic) domain to preserve inter-channel color correlations, which are typically lost in grayscale or conventional RGB processing. To address the ablation study, we have expanded Section 6 to explicitly compare: (1) Baseline YOLOv8, (2) YOLOv8 + Conventional enhancement (CLAHE/Wavelets), and (3) our proposed YOLOv8 + QGF. Comment 3: The manuscript should provide a fully reproducible experimental protocol and i suggest to include the exact train-validation-test split, final preprocessing parameters, fixed YOLO v8 hyperparameters, augmentation settings, inference thresholds. Response 3: We have clarified that the novelty lies in the use of a hypercomplex (quaternionic) domain to preserve inter-channel color correlations, which are typically lost in grayscale or conventional RGB processing. To address the ablation study, we have expanded Section 6 to explicitly compare: (1) Baseline YOLOv8, (2) YOLOv8 + Conventional enhancement (CLAHE/Wavelets), and (3) our proposed YOLOv8 + QGF. Comment 4: The authors should carefully revise the manuscript to resolve inconsistencies in the reported performance metrics especially the discrepancy between the mAP values reported in different sections of the paper. Response4: We have synchronized all metrics. The discrepancy between the 12.7% improvement initially mentioned and the results in Table 3 was corrected. The manuscript now consistently reports a 2.07% mAP increase over the baseline YOLOv8 and justifies the previously mentioned higher percentages as comparisons against non-quaternionic methods like CLAHE (6.2%) and Wavelets (8.9%). Comment 5: The Results section should be strengthened with more rigorous quantitative evaluation under identical experimental conditions. And include baseline comparisons, repeated runs, standard deviations and a clearer error analysis of false positives and false negatives. Response5: Table 3 has been updated to include the mean values and standard deviations from 5 independent runs to ensure statistical stability. We also added a deeper error analysis focusing on how QGF reduces false positives by 18% by mitigating specular reflections. Comment 6: The manuscript requires substantial language editing and structural revision. Several sections are currently written in a speculative or incomplete manner and some figure captions and subsections are not presented in a publication ready form. Response 6: The manuscript has undergone professional English editing. Figure captions (e.g., Figures 6 and 7) were rewritten to describe scientific observations (textural enhancement) rather than software operational steps. Comment 7: The authors should reduce the excessive theoretical background and devote more space to the practical implementation details and experimental validation of the proposed method. Response7: We have streamlined Sections 2.1 to 2.3, removing general quaternionic theory to focus exclusively on the mathematical application to Gabor filters. This space was reallocated to Sections 3 and 4 for implementation details. Comment 8: The limitations imposed by the relatively small and imbalanced dataset should be discussed more explicitly and the claims regarding robustness and generalization should be moderated accordingly or supported by additional external validation.   Response8: We have added a "Limitations" paragraph in Section 5. While acknowledging the 284 images is a relatively small sample, we highlight the density of the dataset with 5,807 whitefly instances and have moderated our claims about universal generalization.   Comment 9: Several strong claims should either be supported by direct experimental evidence and statistical justification or be rewritten in a more cautious and scientifically grounded manner. Response 9: Strong claims were rewritten with a more cautious tone. The Conclusions section was revised to state that the proposed method "shows potential for greenhouse monitoring" instead of claiming "universal robustness," strictly aligning with the evidence presented in Table 3. The conclusion section should be revised so that its claims remain strictly aligned with the results actually demonstrated in the study, without extending beyond the presented evidence.   Regarding the article titled: We greatly appreciate the time and effort you have dedicated to reviewing our work. We have carefully considered all of your recommendations. In response to your observations, we have thoroughly revised the manuscript: we strengthened the experimental basis, corrected the comparative analysis (focusing on our own ablation study rather than direct literature comparisons), improved the quality of all figures and illustrative material, and carefully reviewed and corrected the presentation of mathematical expressions. We believe these changes have substantially improved the clarity and rigor of the paper. We are deeply grateful for your valuable feedback, which has helped us enhance the quality of our work. Thank you again for your respectful and insightful review.   Sincerely, The Authors

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors responded constructively to the criticism, carefully considered most of the comments, and made significant changes to the text of the article. When formatting the final version, please consider the following technical issues: It is necessary to check that the tables are mentioned with the correct numbering: «Table ??» .

Author Response

We would like to thank the reviewer for their constructive feedback and for acknowledging the significant improvements made to the manuscript. We have carefully addressed the remaining technical issues to ensure the paper meets the journal's publication standards

Comments 1: The authors responded constructively to the criticism, carefully considered most of the comments, and made significant changes to the text of the article. When formatting the final version, please consider the following technical issues: It is necessary to check that the tables are mentioned with the correct numbering: «Table ??»
Response 1: The manuscript has undergone a thorough language editing process to improve clarity and ensure that the research is expressed more precisely. Sentences have been restructured for better flow and professional academic tone. We have performed a complete audit of the document's cross-references. All instances of "Table ??" have been replaced with the correct numbering. Table 2 is now correctly cited in the comparison of enhancement techniques, and Table 3 is referenced regarding the performance metrics of the YOLOv8+Gabor model.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

The reported performance values should be carefully reconciled across the manuscript. At present, some central claims do not appear to be fully aligned with the values presented in the tables. For example, the reported 12.7% improvement should be explained clearly.

The manuscript still requires substantial English editing, as several expressions remain below publication-ready level. Examples include  “rerproducible,” “demostrate,” “yelloe sticky traps,” “this method is useful like a Figure.10,” “Table ??,” and “techniques ??”.

Author Response

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.

Comments and Suggestions for Authors

Comment 1: The reported performance values should be carefully reconciled across the manuscript. At present, some central claims do not appear to be fully aligned with the values presented in the tables. For example, the reported 12.7% improvement should be explained clearly.

The manuscript still requires substantial English editing, as several expressions remain below publication-ready level. Examples include  “rerproducible,” “demostrate,” “yelloe sticky traps,” “this method is useful like a Figure.10,” “Table ??,” and “techniques ??”.Response1: Thank you for your valuable feedback. We have performed a major revision focusing on methodological clarity, internal consistency, and linguistic quality. We have restructured the experimental section to ensure reproducibility and validated our results with more rigorous statistical evidence.

.

Response 1: We have reconciled all reported values across the Abstract, Results, and Conclusions. We clarify that the 12.7% improvement in mAP@0.5 specifically refers to the performance gain observed under critical lighting conditions (e.g., overexposure and variable shading) compared to raw, unprocessed data. This is distinguished from the 2.07% global improvement over the baseline YOLOv8 model reported in Table 3. We have updated Section 5 and the Conclusions to explicitly state this distinction and ensure the narrative is fully aligned with the tabular data. We apologize for the technical errors in the previous version. The manuscript has undergone a substantial professional English edit. All specific typos identified (e.g., “rerproducible,” “demostrate,” “yelloe sticky traps”) have been corrected. Expressions such as “this method is useful like a Figure 10” have been rewritten in a more scientific and formal academic style (e.g., "As illustrated in Figure 10, the proposed method enhances...").

Round 3

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript is improved overall. however, minor issues remain regarding language quality, internal consistency of reported metrics (particularly PSNR) and some imprecise or weak formulations. The authors are encouraged to perform one final careful revision to ensure that all numerical claims are fully aligned across the text, tables and conclusions and that the wording is consistently clear and publication ready.
For example, some problematic expressions and typographical errors still appear in the current version, such as: “rea¿liable”, “validate tht te”, “Ganor filter whitch”, “is suported by objetive”, “cuaternion”, “nesesary”, “subjetivity”, “vesion”, “performanceof”, “tecniques”, “traslate”, and “greenhous enviroments”.

Author Response

Dear Reviewer,

We would like to express our sincere gratitude for your insightful feedback and the time invested in reviewing our manuscript. Your observations regarding language quality and metric consistency have been fundamental in elevating the professional standard of this work.

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.

Comment 1: The manuscript is improved overall. however, minor issues remain regarding language quality, internal consistency of reported metrics (particularly PSNR) and some imprecise or weak formulations. The authors are encouraged to perform one final careful revision to ensure that all numerical claims are fully aligned across the text, tables and conclusions and that the wording is consistently clear and publication ready.

For example, some problematic expressions and typographical errors still appear in the current version, such as: “rea¿liable”, “validate tht te”, “Ganor filter whitch”, “is suported by objetive”, “cuaternion”, “nesesary”, “subjetivity”, “vesion”, “performanceof”, “tecniques”, “traslate”, and “greenhous enviroments”.

.

Response 1: Following your recommendations, we have performed a final comprehensive revision.

Below is a point-by-point description of the corrections made:

1. Language Quality and Typographical Errors

We have conducted a thorough manual and automated proofreading of the entire manuscript. Specifically, all the typographical errors highlighted in your report have been corrected as follows:

• "rea¿liable" - reliable
• "validate tht te" - validate that the
• "Ganor filter whitch" - Gabor filter which
• "is suported by objetive" - is supported by objective
• "cuaternion" - quaternion/quaternionic
• "nesesary" - necessary
• "subjetivity" - subjectivity
• "vesion" - vision/version
• "performanceof" - performance of
• "tecniques" - techniques
• "traslate" - translate
• "greenhous enviroments" - greenhouse environments

2. Internal Consistency of Metrics

As per your suggestion, we have synchronized all numerical claims throughout the text, tables, and conclusions to ensure total alignment: PSNR: Standardized to 34.10 dB across all sections to reflect structural integrity accurately, resolving the previous 4.10 dB inconsistency, mAP@0.5: Unified to 0.950 in the Abstract, Table 3, and Conclusions to match the final experimental results, Time Metrics: We have clearly distinguished between the GPU inference time ($10.3\text{ ms}$) and the total processing time (2.3 s) to explain system latency without ambiguity.

3. Reformulation of Weak Expressions

The Abstract, Discussion, and Conclusions have been rewritten to eliminate imprecise language. We replaced vague descriptions with technical, assertive formulations that better highlight the novelty of the quaternionic approach and its impact on pest detection.

We believe these changes have significantly strengthened the manuscript and made it ready for publication.

Sincerely,

The Authors

Author Response File: Author Response.docx