Review Reports
- Jing Han1,2,
- Junxian Guo1,2,* and
- Jun Zhou1,2
- et al.
Reviewer 1: Anonymous Reviewer 2: Joaquín Guillermo Ramírez-Gil Reviewer 3: Anonymous
Round 1
Reviewer 1 Report
1) Why authors opted for PLS based regression in this work? If the main reason is its inherent capacity to effectively address issues related to multicollinearity, then more detailed explanation on this need to be provided.
2)Authors mentioned SVM efficiently overcome the common neural network challenges of convergence and solution instability. Can you include more explanation on this with respect to the proposed work.
3)Data preprocessing is a crucial step. Details about the dataset used and data pre-processing strategies involved should be mentioned.
4)In some sentences, superscripts are not correctly provided. Need to correct this.
5)The abstract should clearly emphasize the novelty of this work.
6)One of the major concern while using dimensionality reduction algorithm is its adverse impact on accuracy as some important features may be missing. However, authors mentioned that utilization of siPLS, CARS, and SPA algorithms throughout the data reduction procedure guaranteed the preservation of the most informative wavelengths while simultaneously reducing the dimensionality of the data. Although this statement is provided in the manuscript, how these algorithms aid in preservation of the most informative wavelengths is not described. Please include the details in the manuscript.
The quality of English language used in the manuscript is acceptable.
Author Response
Please see the attachment.
Reviewer 2 Report
Dear editor and authors in the following text I present my critical review of the manuscript named: The rapid detection of trash content in seed cotton by near-infrared spectroscopy combined with characteristic wavelength selection.
Abstract and Introduction
The summary and introduction provide in part the foundation for understanding the research's context and objectives. However, improving the transition between the introduction section would enhance the flow of the article. Additionally, providing more detailed information on the advantages of FT-NIR spectroscopy and its potential impact on cotton quality assessment would further engage the reader and emphasize the research's significance. I suggest review the topics published in the last papers: DOI: 10.1094/PDIS-08-21-1774-RE, https://doi.org/10.1016/j.rse.2012.09.019
1. The introduction begins with background information about cotton production in China and the importance of Xinjiang in this context. However, the transition from this background information to the specific research topic could be smoother.
2. The introduction mentions various image-based methods and algorithms used in previous studies for detecting cotton impurities. This demonstrates the existing research landscape and the need for alternative detection methods. In addition, discusses the use of spectroscopic detection technology in agriculture, which sets the stage for the study's focus on FT-NIR spectroscopy. However, it could provide more specific details about the advantages and applications of FT-NIR spectroscopy in this context.
3. The introduction successfully outlines the primary objective of the study: to develop an alternative means of detecting traditional trash in seed cotton. However, it could be more explicit about the research's potential benefits and implications for cotton quality assessment, but the transition from the introduction to the methodology section is abrupt. It would be helpful to provide a smoother transition that explicitly introduces FT-NIR spectroscopy as the chosen method for trash content analysis.
Materials and methods
The Materials and Methods section provides detailed in part information about the experimental procedures and data analysis techniques used in the study, but further clarification of variables, algorithms, and rationale behind method choices would enhance the reader's understanding and confidence in the study's methodology. Here is a critical review of this section with suggestions for improvement:
1. The source of seed cotton samples is mentioned, which is essential for reproducibility. However, more details about the specific location, time of collection, and any potential variations in the samples should be provided.
2. While it's mentioned that 152 samples were collected, it would be helpful to clarify how these samples were distributed among the different algorithms and models to avoid any bias.
3. Equation (1) is presented for calculating trash content, but there is no explanation of the variables used in the equation. A brief description of each variable (Z, ms, mc) and their units should be included for clarity.
4. It's important to provide a reference or source for the established test method used for determining trash content to ensure transparency and reproducibility.
5. The description of the FT-NIR data acquisition process is clear and provides details about the equipment used. However, specifying the specific model of the Thermo Fisher Scientific Corporation's Antaris II NIR spectrometer would enhance clarity.
6. The information about the number of scans and repetitions is provided, but it would be beneficial to explain the rationale behind these settings and how they contribute to data quality.
7. The rationale for spectral variable selection is well-described, emphasizing the need to reduce model complexity. However, it would be helpful to explain how selecting essential characteristic wavelengths improves model accuracy.
8. While siPLS, CARS, and SPA methods are mentioned, more information about how these algorithms work, including their advantages and potential limitations, should be included.
9. The section on PLS and SVM-based regression provides a good overview of the techniques used. However, a brief explanation of why these specific methods were chosen for this study and their relevance to the research objectives would be informative.
10. The explanation of the RBF kernel function for SVM is clear, but it could be improved by providing the formula for the Radial Basis Function (RBF) kernel function (Equation (3)).
11. Additional details about the parameter optimization process for SVM, such as the range of values considered for optimization and the convergence criteria, would enhance transparency.
12. The description of model evaluation metrics (RMSE, R2, RPD) is provided, but it would be beneficial to include the mathematical expressions for these metrics (as done for RMSE) to help readers understand how they are calculated.
13. It's important to explain the significance of the selected performance thresholds for RPD (e.g., RPD > 2.0 indicating a reliable model) and why these thresholds were chosen.
14. The mention of using MATLAB R2021a for modeling is appropriate. However, it would be helpful to specify the relevant toolboxes or functions used within MATLAB for these analyses.
15. In summary, while the Materials and Methods section provides a comprehensive overview of the experimental procedures and data analysis techniques, further clarification of variables, algorithms, and rationale behind method choices would enhance the reader's understanding and confidence in the study's methodology.
Results and discussion
The Results and Discussion sections contain valuable information, improving clarity, organization, and providing additional context would enhance the reader's understanding of your research. Additionally, addressing the specific points mentioned above would contribute to the overall quality of the manuscript.
1. Table 1: This table presents the results of subinterval optimization, which is crucial for the study. However, it lacks detailed descriptions and units for the values presented. It's essential to provide a clear explanation of what these values represent, such as what "RMSECV" stands for, and include the units (if applicable) for better clarity.
2. Figures 1 and 2 show RMSECV as a function of the number of PLS components and prediction vs. measurement plots, respectively. While these figures are informative, they lack axis labels, which are essential for understanding the variables being plotted. Adding axis labels and captions for clarity is necessary.
3. Table 2: This table provides results for PLS-based regression using different wavelength selection methods. While the results are comprehensive, the table could be more organized and reader-friendly. Consider adding a brief description of each column's content, including units where applicable. Also, provide a clear indication of which values are considered favorable or optimal, and why.
4. Figures 6, 7, and 8: These figures illustrate the optimization process using BES, GWO, and SSA algorithms. Like the previous comment, these figures need axis labels and captions to explain what they represent. Additionally, a brief discussion of why these optimization methods were chosen and their significance would enhance the reader's understanding.
5. The discussion section should delve deeper into the interpretation of the results. For instance, in Table 2, it's mentioned that "siPLS-SPA-SSA-SVM" is the preferred predictive model, but the reasons behind this preference and the significance of the results should be elaborated upon.
6. When discussing the different models and methods used in the study, it's important to provide a comparative analysis of their strengths and weaknesses. Explain why certain combinations of methods were more effective than others and under what conditions they might be most suitable.
7. Connect the research findings to practical applications and implications. How could these models and methods be applied in real-world scenarios, such as in agriculture or industry? Discuss the potential benefits and limitations of implementing these models.
8. The discussion section could benefit from better organization and clearer subheadings to guide the reader through the different aspects being discussed. This will make it easier for readers to follow your arguments and conclusions.
9. Ensure that you cite relevant studies and literature to support your claims and findings. This helps establish the context and credibility of your research.
I suggest that the autor inprove the Inglish of manuscript with suggestion of a native
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 3 Report
I have reviewed the article titled "The rapid detection of trash content in seed cotton by near-infrared spectroscopy combined with characteristic wavelength selection" by Han et al. This article aimed to investigate the feasibility of using Fourier transform near-infrared spectroscopy (FT-NIR) for the rapid quantitative analysis of trash content in seed cotton.
I would like to express the following moderate concerns about this manuscript.
The abstract is well-written, but it lacks certain key results, such as SEP (Standard Error of Prediction), SECV (Standard Error of Cross-Validation), and RSD (Relative Standard Deviation). Please, additionally, clarify and discuss these parameters in results. I suggest dedicating at least two lines to explaining the limitations of the study to provide readers with a more comprehensive understanding of its scope and applicability.
Lines 62-80. Many repeated phrases and jumbled abbreviations (eg Wu et al. [12] used near-infrared spectroscopy..., Teye et al. [13] used NIR spectroscopy..., Wang et al. [14] used characteristic wavelength-selective NIR spectroscopy..., Li et al [15] used NIR spectroscopy..., Fortier et al [16] obtained NIR spectral..., Azadnia et al [17] used wavelength selection... , etc.) in each of the sentences in which a study is described. The authors must redo these sentences using appropriate connectors, and synthesize the information, or improve the writing.
Section 2.4, titled "Spectral Variable Selection," appears to be incomplete. It lacks citations and does not clearly define the selection criteria used.
Table 1 should be self-explanatory, providing detailed information within the table itself.
The Figure 2 should include units of measurement.
Figure 6, 7, and 8 should be self-explanatory, providing detailed information within the figure itself. The same for Fig. 9 and put Units in axes.
In the Discussion section state and discuss the limitations. It's important to discuss any limitations of the study. Are there any potential sources of bias or error in the data collection or modeling process? Addressing these limitations can help researchers and readers better understand the scope and applicability of the findings.
The authors should also discuss the possibility of overfitting in machine learning methods. Are all models fitting in the same way?
Data Availability: Inquire about the availability of the data used in the study. Is the data accessible to other researchers for validation or further analysis? Open data sharing can enhance the transparency and reproducibility of scientific research.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
Dear authors and editor,
The submitted manuscript demonstrates noteworthy enhancements and provides satisfactory responses to the majority of inquiries and recommendations. However, it is notable that the authors have not undertaken a review of the recommended literature, particularly regarding the selection of informative spectral bands using multiple spectral responses. This is particularly relevant because a significant portion of these responses are non-specific to the studied phenomenon. In light of this, we once again recommend a thorough revision of this approach and its inclusion in both the introduction and discussion sections of the manuscript.
The Ingles is ok
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
The authors have addressed my comments and suggestions, so the article can be accepted on my part.
Author Response
Please see the attachment.
Author Response File: Author Response.docx