Application of Near Infrared Hyperspectral Imaging Technology in Purity Detection of Hybrid Maize
Round 1
Reviewer 1 Report
The submitted research work aims to use hyperspectral Imaging Technology in Purity Detection of Hybrid Maize. I found this paper very interesting where Several technical aspects were nicely implemented and explained sufficiently. Undoubtedly, authors invested huge amount of time and have made a great effort to produce this high-quality of research which is clearly structured and the language used is largely appropriate. I see that this manuscript in its form and level deserves to be accepted for publication in MDPI-AS.
Author Response
Thank you very much for your approval of the content of my article. I will further check the content and format of the manuscript to meet the requirements of the journal publication.
Reviewer 2 Report
Well written manuscript. Please provide the Y axis name in Fig.5. In future, advance feature wavelength selection methods like CARS, IRIV can be tried for better results.
Author Response
Thank you very much for your careful review of my manuscript. I am deeply sorry for the low-level error in Fig.5. The Y axis name of Fig.5 is the Loading coefficient of each principal component, and I have modified it in the manuscript. In the follow-up research, I will try advanced characteristic wavelength selection methods such as CARS and IRIV, and compare the results to get the best method. I will also further check and revise the content and format of the manuscript to meet the requirements of journal publication.
Reviewer 3 Report
OVERVIEW
The paper describes a method for rapid and non-destructive detection of seed perity. The method uses near-infrared hyperspectral imaging technology for variety identification 23 and purity detection of maize seeds. Its contribution lies in employing SVM as a pattern recognition method for 14 classes of the five different kinds of maize seeds. The method also achieved a high recognition accuracy.
The background details of the manuscript is beyond my expertise. My key comments related to the cross-disciplinary relevance of this work within the area of the ML method.
Please note that I see the peer-review process as a chance to have an academic dialogue between colleagues with different backgrounds and priorities. Do not feel obligated to make changes in response to each of my comments, especially if you disagree with some of them.
GENERAL COMMENTS
(1) The study utilizes SVM for modeling, however, as far as I know, SVM is not a state-of-the-art machine learning method. Do you have any theoretical basis for choosing SVM? Here I am referring to the theoretical aspect of SVM to explain why you used SVM?
(2) Additionally, why don't you consider XGBOOST, which is regarded as one of the best machine learning methods under a small dataset. And algorithms such as RF, knn, etc.Since your dataset is small (100+50), I would doubt the credibility of the results.
(3) You have a thin introductory paragraph, hence I was unable to learn about the current status of research and problems related to seed NDT, classification, purity testing, etc. I suggest you re-organize the related work in your introduction.
(4) For your dataset, this is not a convincing result for a too small dataset in my opinion. I strongly suggest that you can expand your dataset.
(5) There are several grammar errors and typo errors in the thesis. It is suggested to conduct a careful proofreading.
Author Response
Thank you very much for carefully reviewing my manuscript in your busy schedule and putting forward your valuable comments. Your comments are very helpful for the revision of my manuscript and subsequent research！
I have responsed the comments point by point and revised the manuscript, especially the introduction. Please refer to the attachment for details. I also checked and revised the language and format of the manuscript.
Sincerely hope that you can put forward your valuable suggestions again！
Author Response File: 
Author Response.docx
Reviewer 4 Report
This is an interesting manuscript for which only positive feedback should be provided. The article is written with a highly scientific style and is well-organized. The provided information's relevant and worthy of investigation.
A support vector machine with several kernel functions was used to determine the purity of hybrid maize by near-infrared hyperspectral.
The methods are presented in a clear manner, and the obtained results are deeply analyzed and discussed; the discussion section is informative, and finally, the figures and tables are of good quality.
I have read this manuscript and only have several parts for little amendments, especially in the chapter of Introduction.
# Please answer and put your introduction chapter in one paragraph again (cite from another reference if needed) to clarify your research gap.
"Why this research chose the SVM algorithm and just only one algorithm from machine learning if the conventional algorithm PLSR can give you the best model?"
"What basic theory is behind that?" The author should not just write a previous paper but must know the basic theory behind that.
Please explain using basic theory.
Author Response
Thank you very much for carefully reviewing my manuscript in your busy schedule and putting forward your valuable comments. Your comments are very helpful for the revision of my manuscript and subsequent research！
I have responsed the comments point by point and revised the manuscript, especially the introduction. Please refer to the attachment for details. I also checked and revised the language and format of the manuscript.
Sincerely hope that you can put forward your valuable suggestions again！
Author Response File: Author Response.docx
Round 2
Reviewer 3 Report
1. Please clarify the size of your experimental dataset, and the exact number of each category. This is critical for classification problems.
2. I don't see your modification in the manuscript about the comparison with other machine learning algorithms. You can show your choice (SVM) by specific theory or by experimental comparison to show that SVM is the most suitable classification algorithm for your task.
I am not an expert in this field, so please refer to my comments as appropriate. Thank you.
Author Response
Dear Reviewer,
Thank you for your comments concerning our manuscript, these comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our researches. We have studied comments carefully and have made correction which we hope meet with approval. We have carefully answered and revised your question, and the contents are as follows:
Point 1: Please clarify the size of your experimental dataset, and the exact number of each category. This is critical for classification problems.
Response 1: Five maize varieties were used in the experiment, and the number of samples for each variety was 150. The training set and test set were divided according to the ratio of 2:1. The specific division can be seen from Table 2. The sample division method is based on Kennard-Stone algorithm, which can ensure that the samples in the training set are uniformly distributed in the spatial distance.
Point 2: I don't see your modification in the manuscript about the comparison with other machine learning algorithms. You can show your choice (SVM) by specific theory or by experimental comparison to show that SVM is the most suitable classification algorithm for your task.
Response 2: We have added the analysis data of PLS-DA, KNN, NB and RF algorithms. Through the comparison of experimental results, it is concluded that SVM is most suitable algorithm. We have mainly revised parts 2.3 and 3.4 of the manuscript, added the principle, characteristics and analysis results of the algorithm. For specific modifications, please see the latest revised version, the revised content has been marked in red.
Once again, thank you very much for your comments and suggestions.
Author Response File: Author Response.docx
