Advancing Visible Spectroscopy through Integrated Machine Learning and Image Processing Techniques

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This work proposed a novel algorithm for visible spectroscopy image processing. Based on the claimed performance, it appears that this algorithm works well. It showed remarkably low error percentages of 0.04%, 0.01%, 3.7%, 1%, and 0.07% for sodium vapour, neon lamp, copper vapour laser, and helium vapour. In general, this work is suitable for applied sciences. Yet, some improvements must be made before its final publication.

(1)    If we look at Figure 5, the polynomial fitting, based on my own fitting experiences, is not good. The fitting and the data trace did not overlap very well. Please provide a reason. Could a 7-order fitting improve the result?

(2)    The data presented in the main text and the experimental details are very thin. It is difficult for the readers to look into the details. I could not find any original data in the manuscript. All I can see is the result from the authors and the authors just told the readers that their fitting performances are good. This is not very convincing. In the revised manuscript, the authors must provide more details.

 

Comments on the Quality of English Language

English is fine.

Author Response

Dear Reviewer

Kindly find the attachment where we have tried to answer all your raised concerns.

Thank you

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors,

The manuscript presents an interesting method for analyzing visible spectral data. While the Authors developed the models correctly, several missing topics can be improved. Below are my comments:

Line 39. Can the authors be more specific about the type of machine learning model used? Terminology in ML can be broad, and different approaches have different strengths and weaknesses.

Lines 65-144. This paragraph is quite long (145 lines) and might benefit from being broken down into smaller sections for improved readability. One suggestion is to divide the paragraph into multiple sections based on the topics or ideas it covers.

Line 221. Can the Authors indicate the company information for this device?

How many samples (data) were used in this study? A table with the number of datasets for each light source could be informative.

Figure 2. This figure is critical in your study. Based on its elements, can the authors describe in detail how the dataset was collected? A table with the different values of the polynomial regression could be helpful. How was calculated the hue value from an image?

The Authors mentioned using the method from previous studies. However, can the Authors explain in detail how the data is acquired and used to develop the models and how the results benefit the scientific community? It is not clear from the manuscript.

I hope my comments help.

Author Response

Dear Reviewer

Kindly find the attachment where we have tried to answer all your raised concerns.

Thank you

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The paper presents a method for visible spectroscopy that combines image processing techniques with machine learning algorithms. The process starts by calculating the hue of an image to determine the dominant wavelength. A 6-degree polynomial regression model is then trained to map the hue values to the dominant wavelengths. This machine learning model is used to analyze the visible wavelengths from different light sources, such as sodium, neon, mercury, copper vapour lasers, and helium vapours. The accuracy is confirmed by error analysis, showing low error rates for each light source. This technique promises to enhance visible spectroscopy’s precision and efficiency. This is a useful study. However, there are several issues that need to be clarified before it is recommended for publication.

1.         The authors should provide in the abstract what machine learning algorithms were used, as well as some qualitative conclusions.

2.         In Section 2, the authors reviewed research in related fields. The author should categorize these studies and provide a brief summary of each category. Otherwise, readers would be very confused to see these studies. This listing is very bad.

3.         The description of the instrument should be based on specific performance indicators, rather than the authors commenting on how advanced it is.

4.         For machine learning models, specific descriptions are also provided in non scientific language. It seems that the description is automatically generated using a big data model.

5.         In Figure 1, if the authors want to provide the code, simply provide the text without the need for screenshots.

6.         For “In this innovative study, machine learning algorithms were seamlessly integrated into an image processing framework”, the “innovative study”, this description is extremely inappropriate. There are many similar descriptions in the manuscript, which should be avoided. Others, such as: “ novel”, “the overall performance of the methodology remains commendable.”

7.         Why did authors choose polynomial regression models?

8.         For “One notable technique within machine learning is polynomial regression, which is a type of regression analysis where the relationship between the independent variable (or variables) and the dependent variable is modelled as an nth-degree polynomial function.” This description is unreasonable as there are many regression models. Not just polynomial regression.

9.         Why did not authors use their own experimental data when citing published image data.

10.     What are the evaluations of model performance based on? How was it calculated and obtained?

Author Response

Dear Reviewer

Kindly find the attachment where we have tried to answer all your raised concerns.

Thank you

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The authors have developed a method to visible spectroscopy by using image processing techniques and machine learning algorithms. However, the manuscript is  poorly written and it is challenging to ascertain the novelty of the research work. Significant revisions are required for the manuscript before it can be further considered for possible publication.

1.      The “literature review” includes a long list of works, which is both difficult to navigate and lacks clear organization. To improve the presentation, it is important to restructure and summarize the content in a more systematic way. It is also important to emphasize the existing applications or the need for using image processing and machine learning in spectroscopy.

2.      The “methodology” contains an excess of information that is either already known or expected to be known by readers. Furthermore, there are repeated emphases on the benefits of image processing and machine learning. It is recommended to streamline the introduction by avoiding repetitive statements and focusing on providing concise explanations.

3.      The authors state in the manuscript that the dataset used was obtained from reference [21]. However, reference [21] does not contain any data on hue and wavelength. Therefore, please provide the correct reference paper for the dataset used in the study.

4.      In line 285, the authors state that “captures the complex relationship between hue values and image hues”, what does this mean?

5.      This work examines the connection between hue and wavelength through the utilization of machine learning for polynomial training. The model that is trained is subsequently verified using the wavelengths emitted by five distinct vapor light sources. Based on the results shown in the figure, it is evident that the relationship between hue and wavelength is relatively simple, enabling a direct polynomial fit to analyze their correlation. Please contrast the disparities between using machine learning for polynomial training and fitting a polynomial directly, thus emphasizing the importance of employing machine learning.

 

Comments on the Quality of English Language

NA

Author Response

Dear Reviewer

Kindly find the attachment where we have tried to answer all your raised concerns.

Thank you

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The current manuscript was indeed improved compared with previous one. Yet, certain improvements must be made to meet a decent publication standard of a scientific paper.

(1)    Figure 4, 5, 6 should be prepared according to the same style and the inset should be within (rather than outside) the main figure.

(2)    As the authored mentioned that “Regarding the suggestion for a 7-order fitting, experimentation showed that it did not improve the result; in fact, it resulted in poorer fitting.”, the results should be included in the manuscript, e.g. in Figure 4.

Author Response

Kindly find the attachment for the reply.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have made significant positive revisions to the draft. I believe the current format is now suitable for publication.

Author Response

Kindly find the attachment for the reply.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

 

The "literature review" sectionis quite lengthy, spanning nearly 4 pages. I recommend condensing it to a shorter lengh of 1-1.5 pages.

  Comments on the Quality of English Language

No

Author Response

Kindly find the attachment for the reply.

Author Response File: Author Response.pdf