Using RGB Imaging, Optimized Three-Band Spectral Indices, and a Decision Tree Model to Assess Orange Fruit Quality
Round 1
Reviewer 1 Report
The authors investigated the use of 1) invasive sampling, 2) new three-band reflectance indices, 3) traditional reflectance indices and 4) simple RGB indices, the reflectance indices and the combination of these as an input to a decision three model to estimate the ripeness of Navel oranges. The approach is simple and robust but the methodology of this study is poorly described. Writing style is confusing in many places and the central content requires crystallization. Manuscript requires a thorough proofreading.
Revise the abstract to clearly state the objectives and the main results; see what you say in the end of the Introduction - this was clearly written. Revise also the conclusions to maintain the same logical order of handling the results. The final conclusion of a possible active sensor system is interesting and well-formulated - could this be stated already in the introduction? Consider also revising the title to better describe the work: maybe instead of "integration" the authors might simply say "use"?
The introduction is generally well written, but the knowledge gap remains unclear. Please note that this is a practical application, so I would avoid formulating larger (artificial) knowledge gap, as repetition has value in science, too. Alternatively the authors might present here the idea of an in situ sensorics.
Figure 2: "Waveband" should be used instead of "wavelenght" as these bands have a certain width. Why the authors display regression coefficients instead of correlation coefficients? This should be reasoned.
Figure 3 shows a flow chart of this study, however, it actually describes only part of the objective iv stated in the introduction.
Collection of plant material is well described and Figure 1 is informative. Chapters 2.2. and 2.3. require more details (measurement conditions, sample volumes). Give also information of the light source for the RGB imaging and spectral reflectance measurements (sunlight?).
R16: "entirely" is not correct here as the authors clarify later
R17: Remove "incredibly"
R18-23: The aim is not clearly stated; consider dividing this to several sentences.
R24: Remove "significantly" if not followed by metrics of statistical signifigance
R25: More successful than what?
R33-35: The model-related acronyms, such as DT-SRIs -CI-30, have not been explained.
R43: Check the formatting of the FAO -reference.
R46: Check reference formatting
R62: What do you mean by "sickness" here?
R70: Revise the terminology here, as "spectroscopy" can be considered as a top concept for "hyperspectral reflectance" and "hyperspectral imaging"
R144-145: Consider revising the knowledge gap. This does not logically follow from what has been written above.
Table 1 has not referred to in the text.
R192: Change "shot" to "imaged"
R196: Explain "overcast settings"
R201: please give an example of "non-fruit features"
R201-202: This has already been said earlier
R228: Change "is" to "was"
R231: Modify how?
R265: Vague statement, consider omitting or clarifying.
R314: Delete "such as"
R408: Compare your results to others', not vice versa as done here.
R413: Explain firmness. How and when this was measured?
R418: Delete "extreme substantially"
R455: Explain these model-related acronyms properly. Note here that both L* and L are used in the text.
Revise the Data Availability Statement - there are no original measurement data given in the article; only processed results.
Acknowledgements repeat what has been said in the Funding statement.
Author Response
Response to Reviewer #1 Comments
Reviewer #1:
We greatly appreciate your critical observations as well as your constructive and helpful comments. We hope that we could address your questions/comments by the explanations and revisions made in the manuscript.
- Comment: Revise the abstract to clearly state the objectives and the main results; see what you say in the end of the Introduction - this was clearly written.
Response:
Thank you for your guidance and support for our manuscript. Already, we have rewritten the abstract section in lines 15-37. We hope this modification is acceptable to you.
- Comment: Revise also the conclusions to maintain the same logical order of handling the results. The final conclusion of a possible active sensor system is interesting and well-formulated - could this be stated already in the introduction?
Response: Thank you for your comments. We have followed your suggestions to present the paper clearly. Therefore, we revised the conclusion section again and described this section clearly in the lines 498-515.
- Comment: Consider also revising the title to better describe the work: maybe instead of "integration" the authors might simply say "use"?
Response: Thank you for your comments. We revised the title again to be “Using RGB Imaging, Optimized Three-Band Spectral Indices and a Decision Tree Model to assess Orange Fruits Quality”.
- Comment: The introduction is generally well written, but the knowledge gap remains unclear. Please note that this is a practical application, so I would avoid formulating larger (artificial) knowledge gap, as repetition has value in science, too. Alternatively the authors might present here the idea of an in situ sensorics.
Response: Thank you for your good comments. We reviewed the knowledge gap during the introduction again in lines 133-145.
- Comment: Figure 2: "Waveband" should be used instead of "wavelength" as these bands have a certain width. Why the authors display regression coefficients instead of correlation coefficients? This should be reasoned.
Response: We are sorry for this error. So, we replaced these words “wavelength or regression coefficients” and made them right throughout the paper.
- Comment: Figure 3 shows a flow chart of this study, however, it actually describes only part of the objective IV stated in the introduction.
Response: We apologize for the inappropriate description. We examined this sentence again and changed it according to your suggestions in line 287-288.
- Comment: Collection of plant material is well described and Figure 1 is informative. Chapters 2.2. and 2.3. require more details (measurement conditions, sample volumes). Give also information of the light source for the RGB imaging and spectral reflectance measurements (sunlight?).
Response: Thank you for your good comments. Measurements condition and light source for RGB was modified from line 193 to 195. As well as for spectral reflectance measurements were modified from line 227 to 229.
- Comment: R16: "entirely" is not correct here as the authors clarify later
Response: We apologize that we didn't explain this expression clearly. We examined this sentence again and changed it according to your suggestions in lines 15-16. We hope this expression could be agreeable to you.
- Comment: R17: Remove "incredibly".
Response: Thank you for your good comments. We have deleted this word and rewritten this sentence again in lines 16-17.
- Comment: R18-23: The aim is not clearly stated; consider dividing this to several sentences.
Response: Thank you for your valuable suggestions. We have modified the writing according to your comments in lines 17-23.
- Comment: R24: Remove "significantly" if not followed by metrics of statistical signifigance.
Response: Thank you for your comments. Already, we have omitted this word.
- Comment: R25: More successful than what?
Response: Thank you for your positive comments. We have written this sentence again in lines 25-26.
- Comment: R33-35: The model-related acronyms, such as DT-SRIs -CI-30, have not been explained.
Response: Thank you for your valuable suggestions. We have revised this sentence again in lines 30-34. We hope this expression could be agreeable to you.
- Comment: R43: Check the formatting of the FAO -reference.
Response: Thank you. We deleted this word from our writing (FOA, 2022). Since it is sufficient to refer to the reference using [1].
- Comment: R46: Check reference formatting.
Response: Thank you. We removed this part from our writing “(Ministry of Agriculture and Land Reclamation, 2019)”. Since it is sufficient to refer to the reference using [2].
- Comment: R62: What do you mean by "sickness" here?
Response: Thank you. We used another word in line 60.
- Comment: R70: Revise the terminology here, as "spectroscopy" can be considered as a top concept for "hyperspectral reflectance" and "hyperspectral imaging"
Response: Thank you. We have rewritten this sentence again in lines 69-70.
- Comment: R144-145: Consider revising the knowledge gap. This does not logically follow from what has been written above.
Response: Thank you. We have omitted this sentence and revised the writing in lines 133-145.
- Comment: Table 1 has not referred to in the text.
Response: Thank you. We referred this table in our writing in lines 212-213.
- Comment: R192: Change "shot" to "imaged"
Response: Thank you. In line 190, we replaced this word to be “imaged”.
- Comment: R196: Explain "overcast settings"
Response: Thank you. In line 194, we replaced this expression to be “cloudy conditions”.
- Comment: R201: please give an example of "non-fruit features"
Response: Thank you. In line 198-199, we have given an example of non-fruit features.
- Comment: R201-202: This has already been said earlier
Response: Thank you. Already, we deleted this sentence from our writing.
- Comment: R228: Change "is" to "was"
Response: Thank you. In line 255, we have changed this word to be “was”.
- Comment: R231: Modify how?
Response: Many thanks for this comment. it was corrected in the line 230.
- Comment: R265: Vague statement, consider omitting or clarifying.
Response: Thank you. In line 169-170, we have rewritten this sentence again to be clear.
- Comment: R314: Delete "such as"
Response: Thank you. In line 320-321, we have omitted this word in our writing.
- Comment: R408: Compare your results to others', not vice versa as done here.
Response: Thank you. In line 414-417, we have rewritten this sentence in our writing again.
- Comment: R413: Explain firmness. How and when this was measured?
Response: It was explainED from line 422 to 424.
- Comment: R418: Delete "extreme substantially"
Response: Thank you for your suggestions. Already, we deleted this description from our writing in lines 424-425.
- Comment: R455: Explain these model-related acronyms properly. Note here that both L* and L are used in the text.
Response: Thank you for your comments. We reviewed these acronyms again in lines 461-464.
- Comment: Revise the Data Availability Statement - there are no original measurement data given in the article; only processed results.
Response: Thank you for your suggestions. We have modified according to your valuable comments in line 525.
- Comment: Acknowledgements repeat what has been said in the Funding statement.
Response: Thank you for your suggestions. We have changed this description according to your valued observations in line 526-527.
Author Response File: 
Author Response.docx
Reviewer 2 Report
This paper presents a technique to access the orange fruit quality using camera systems and machine learning (Decision tree). The article is well written and provides promising results. However, some minor revisions could be addressed to improve the paper, such as:
1. Since the camera system is employed, the lighting problems should be discussed. Authors have written how to take data images for the experiments. But it seems too tight for real implementation.
2. In the work, fruit is classified into mature, semi-ripening, and ripening. How to define the category is not clearly explained.
Author Response
Response to Reviewer 2 Comments
Reviewer #2:
We greatly appreciate your critical observations as well as your constructive and helpful comments. We hope that we could address your questions/comments by the explanations and revisions made in the manuscript.
- Comment: Since the camera system is employed, the lighting problems should be discussed. Authors have written how to take data images for the experiments. But it seems too tight for real implementation.
Response:
Thank you for your good comments. It was explained from line 193 to 195. The measurements were done in close door under cloudy conditions to prevent the light problems. The measurements were carried out under cloudy conditions to guarantee high image resolution. The flash of the camera was always kept off during measurements.
- Comment: In the work, fruit is classified into mature, semi-ripening, and ripening. How to define the category is not clearly explained.
Response: Thank you for your good comments. According to the peel color photo and it was explained under section 2.2.2. Peel Color.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The authors have improved and clarified manuscript, but the following parts still require inspection:
R18: The authors added "a valuable promising implement" in the abstract. Replace this by a more accurate statement, e.g. "a non-invasive protocol".
R69: Remove this newly added sentence as spectroscopy is not an unconventional technique per se.
R161: Format so that the sentence does not start with a number.
R239-241: Add here how the best indices were selected. I guess they were not just selected from the Figure 2 by eyeballing?
Figure 2: The authors have not fully responded to my question of why regression coefficients (R2) were shown instead of correlation coefficients (e.g. pearson correlation, r?) in the correlograms. Was this to avoid negative correlations in Figure 2? Or is this to facilitate comparison between the index-based method and the DT-modeling results? If so, please elaborate. Correlograms in Figure 2 contain information that is not shown to the reader (the two sides and the furthermost corner). Please note this in the figure legend.
R270-271: Remove this new sentence as it is not grammatically correct and does not provide new information.
R282: Please, consider elaborating the selection of hyperparameters in more details. What did "a loop" mean here? This study shows that the model performed well, so hyperparameter tuning is not the most crucial part here. However, I would like to see this clarified.
Data availability: Consider publishing the original research data, and the Python codes of the models (e.g. as the appendix) to increase transparency.
Author Response
Response to Reviewer 2 Comments
We greatly appreciate your critical observations as well as your constructive and helpful comments. We hope that we could address your questions/comments by the explanations and revisions made in the manuscript.
- Comment: R18: The authors added "a valuable promising implement" in the abstract. Replace this by a more accurate statement, e.g. "a non-invasive protocol".
Response: Thanks for your positive comment. We have replaced this statement with "non-invasive protocol" in line 18. We hope this modification is acceptable to you.
- Comment: R69: Remove this newly added sentence as spectroscopy is not an unconventional technique per se.
Response: Thank you for valuable comment. We have updated this description with one that's more precise in lines 70-71.
- Comment: R161: Format so that the sentence does not start with a number.
Response: We apologize for this mistake. This sentence has been rewritten to be clear to readers.
- Comment: R239-241: Add here how the best indices were selected. I guess they were not just selected from the Figure 2 by eyeballing? Figure 2: The authors have not fully responded to my question of why regression coefficients (R2) were shown instead of correlation coefficients (e.g. pearson correlation, r?) in the correlograms. Was this to avoid negative correlations in Figure 2? Or is this to facilitate comparison between the index-based method and the DT-modeling results? If so, please elaborate. Correlograms in Figure 2 contain information that is not shown to the reader (the two sides and the furthermost corner). Please note this in the figure legend.
Response: Thanks for your suggestions. 3-D maps in this work displaying only the coefficient of determination (R2) between newly three SRIs and fruit quality parameters. We used the python code for select the best indices constructed by combined three wavelengths. We have added some important information on how to choose the best indices based on the coefficient of determination in lines 241-250.
- Comment: R270-271: Remove this new sentence as it is not grammatically correct and does not provide new information.
Response: Thank you for sharing with us to improve this paper. Already, we removed this sentence.
- Comment: R282: Please, consider elaborating the selection of hyperparameters in more details. What did "a loop" mean here? This study shows that the model performed well, so hyperparameter tuning is not the most crucial part here. However, I would like to see this clarified.
Response: Thank you. We mentioned in the introduction in lines (130-132) "hyper parameter selection has a significant impact on the performance of any machine learning model, which has multiple advantages: it can improve the performance of ML algorithms [38]”. The materials and method section also contained some important information about hyperparameters in lines (279-284) and lines (297-302).
- Comment: Data availability: Consider publishing the original research data, and the Python codes of the models (e.g. as the appendix) to increase transparency.
Response: Thank you. We have attached the Python code as supplementary materials to this paper but original data include very huge data of spectra reflectance from 302 to 1148 nm and we cannot display them in appendix. If the reader needed we can attached the files.
Author Response File: Author Response.docx
