Rapid Nondestructive Detection of Chlorophyll Content in Muskmelon Leaves under Different Light Quality Treatments

Round 1

Reviewer 1 Report

This paper focuses on the "Rapid nondestructive detection of chlorophyll content in muskmelon leaves under different light quality treatments". Although the paper is suitable for publication in this Journal, however, there are several drawbacks that need to be attended to before it is considered again. The following are my suggestions/comments:

There are several types and punctuation errors; these need to be corrected. It is suggested that the ms be checked by an English language editor.

Abstract:

Some details of the experimental method and results are to be included. 

Please mention the name of the authority after Melon.

 

Introduction:

What is the geographical distribution, area of cultivation and production of Melon plants in the country? Please add information.

This section is also long and need to be reduced and written in a precise manner.

The objective of the study is not written properly; the language must be improved so as to reflect its importance.

Materials and methods:

 source of all chemicals, solvents, instruments, etc is to be mentioned. Site details (including geographical coordinates) must be mentioned in brief. What is the size of the controlled environmental room?

The conditions of the experiment? you must write the detail.

The sequence of different steps mentioned (lines 98-106) is unclear and difficult to understand. Please write the various steps from sowing to raising seedlings, followed by transfer to another system and subsequent treatments in a simple manner.

 

Results and Discussion:

The Result and discussion sections are long and can be easily reduced.

 

Conclusions:

In view of the above-mentioned reasons the ms cannot be accepted in its present form and requires a major revision.

please check the attached file.

Comments for author File: Comments.pdf

Author Response

1. There are several types and punctuation errors; these need to be corrected. It is suggested that the ms be checked by an English language editor.

Response: Thanks for the your suggestion. According to advice, we corrected for errors in type and punctuation.

2. Abstract:

Some details of the experimental method and results are to be included. 

Please mention the name of the authority after Melon.

Response: Thanks for the your suggestion. According to advice, we revised the abstract and explained some details of the experimental method and results in the abstract in Line 10-18.

3. Introduction:

What is the geographical distribution, area of cultivation and production of Melon plants in the country? Please add information.

Response: Thanks for the your suggestion. According to advice, we added the geographic distribution, planting area and production area of muskmelon in China in lines 31-38.

This section is also long and need to be reduced and written in a precise manner.

Response: Thanks for the your suggestion. According to advice,we simplify and write this part in a precise way.

The objective of the study is not written properly; the language must be improved so as to reflect its importance.

Response: Thanks for the your suggestion. According to advice, we described the research purpose of this experiment in lines 64-76.

4. Materials and methods:

 source of all chemicals, solvents, instruments, etc is to be mentioned. Site details (including geographical coordinates) must be mentioned in brief. What is the size of the controlled environmental room?

Response: Thanks for the your suggestion. According to advice, in this part, we describe the source of the instrument, and also briefly mention the geographical location of the test site and the size of the controlled environment room. The size of the controlled environmental room is 40 m².

The conditions of the experiment? you must write the detail.

Response: Thanks for the your suggestion. According to advice, we briefly describe the experimental conditions in lines 82-88.

The sequence of different steps mentioned (lines 98-106) is unclear and difficult to understand. Please write the various steps from sowing to raising seedlings, followed by transfer to another system and subsequent treatments in a simple manner.

Response: Thanks for the your suggestion. According to advice, in lines 89-96, we wrote the whole process from sowing to seedling raising, and then transplanting seedlings to the plant factory to the growth and development of muskmelon, and drew the flow chart of the whole growth cycle of muskmelon, as shown in Figure 1.

5. Results and Discussion:

The Result and discussion sections are long and can be easily reduced.

Response: Thanks for the your suggestion. According to advice, we have modified this part.

6. Conclusions:

In view of the above-mentioned reasons the ms cannot be accepted in its present form and requires a major revision.

please check the attached file.

Response: Thanks for the your suggestion. According to advice, we made major revisions to this document and checked the attached documents.

 

Author Response File: Author Response.doc

Reviewer 2 Report

A lot of experimental work focus on rapid nondestructive detection of chlorophyll content has been done in this paper. There are some suggestions that need to be corrected.

1. In section 2.3, I suggest explaining the situation when acquiring hyperspectral images. If possible, it is better to add a sketch map or a photo of the sampling. At the same time, I also have a question: Is there any interference in collecting directly under the fill light of the plant factory?

2. Line 326  Here, the RMSEP value of CNN is 1.273, which is much smaller than the other two models. However, in Table 4, the RMSEP value of CNN is 2.056, so I suggest you check the table carefully.

3. In the conclusion, it is suggested to explain T3, otherwise, the reader needs to return to Table 1 again, which will damage the reader's reading experience.

Author Response

1. In section 2.3, I suggest explaining the situation when acquiring hyperspectral images. If possible, it is better to add a sketch map or a photo of the sampling. At the same time, I also have a question: Is there any interference in collecting directly under the fill light of the plant factory?

Response: Thanks for the your suggestion. According to advice, in 2.3.1, we described the test site for hyperspectral image acquisition, and added the schematic diagram of hyperspectral imaging system, as shown in Figure 2. We also replied that there would be no interference if we collected directly under the fill light of the plant factory. Because the six treatments have different light quality, other environmental parameters are consistent (temperature: 26-35℃in the daytime, 14-18℃ at night; humidity: 65-75%; carbon dioxide concentration: 400-1200 ppm, covering an area of 225 m²). The size of the control environment room is 40 square meters. In addition, the plant fill light (blue light wavelength: 400-499 nm, red light wavelength: 600-700 nm, ultraviolet light wavelength: 380-399 nm and far infrared light wavelength: 701-780 nm) we use is suitable for plant growth and development, which is equivalent to the external natural light. Therefore, there is no interference when collecting directly under the plant fill light.

2. Line 326  Here, the RMSEP value of CNN is 1.273, which is much smaller than the other two models. However, in Table 4, the RMSEP value of CNN is 2.056, so I suggest you check the table carefully.

Response: Thanks for the your suggestion. According to advice, We checked Table 4, reorganized the Rp value and RMSEP value of PLSR in Table 4, and changed 1.273 to 2.055, making it consistent with CNN's RMSEP value in Table 4 and Figure 8-B.

 

3. In the conclusion, it is suggested to explain T3, otherwise, the reader needs to return to Table 1 again, which will damage the reader's reading experience.

Response: Thanks for the your suggestion. According to advice, we explained in the conclusion that the specific proportion of T3 treatment (Light ratio: 6R/1B/2W, Light quantum flux: 360 μmol/(m²·s), Photoperiod: 12 h) in line 281-282.

Author Response File: Author Response.docx

Reviewer 3 Report

With all due respect, the current manuscript is of poor quality and unacceptable for publication. It needs a lot of serious revision.

1. Title. What's the difference between "muskmelon" and "melon"? The descriptions should be consistent in the manuscript.

2. Authors. There are too many authors listed for this uncomplicated paper, it is suggested to put some authors in the acknowledgment. Then, only one corresponding author is allowed for MDPI Journals, and please replace the non-institution email. Moreover, this is just common sense, as "Ling Ma and Yao Zhang contributed equally to this work", they should have be listed as co-first authors, not co-corresponding authors.

3. Abstract. Please rewrite it. This section lists a plethora of methods and experimental procedures without focus. The real necessity, highlights and innovations of this paper were not stated.

4. Introduction. This part of the research progress is not sufficient, and there is no logical relationship between different paragraphs. I don't know what the author wants to express, because it has no summary of the research problems or innovations of this paper. What's more, I'm wondering what's so special about studying melons.

5. Materials and Methods. The paper lacks the introduction of the spectral features, growth environment and other aspects of the experimental material "Boyang 91", and the descriptions of the experimental method are very incomplete.

6. Data processing and analysis. There are no legends in Figure 1 and Figure 2. Then, the selection of pretreatment method, extraction of feature wavelength and comparison of prediction model are all based on only one index, which lacks the reliability of experimental results.

7. Discussion. There is no discussion.

8. Conclusion. There is no prospect based on the research of this paper in the conclusion, and the logic between paragraphs must be improved.

Author Response

1. What's the difference between "muskmelon" and "melon"? The descriptions should be consistent in the manuscript.

Response: Thanks for the your suggestion. According to advice, we use “muskmelon” to describe all muskmelons in the manuscript.

2. There are too many authors listed for this uncomplicated paper, it is suggested to put some authors in the acknowledgment. Then, only one corresponding author is allowed for MDPI Journals, and please replace the non-institution email. Moreover, this is just common sense, as "Ling Ma and Yao Zhang contributed equally to this work", they should have be listed as co-first authors, not co-corresponding authors.

Response: Thanks for the your suggestion. According to advice, we thankedMa Yan, Ma Siyan and Du Minghua, and then replaced the non institutional emails, and listed Ma Ling and Yao Zhang as the co-first authors

3. Please rewrite it. This section lists a plethora of methods and experimental procedures without focus. The real necessity, highlights and innovations of this paper were not stated.

Response: Thanks for the your suggestion. According to advice, we rewrote the abstract and explained its real necessity, focus and innovation.

4. This part of the research progress is not sufficient, and there is no logical relationship between different paragraphs. I don't know what the author wants to express, because it has no summary of the research problems or innovations of this paper. What's more, I'm wondering what's so special about studying melons.

Response: Thanks for the your suggestion. According to advice, we have overhauled this part and linked the logic between each paragraph, and also described the research purpose of this experiment in lines 38-45, 51-52 and 73-76 to reflect the importance of this study.

5. Materials and Methods. The paper lacks the introduction of the spectral features, growth environment and other aspects of the experimental material "Boyang 91", and the descriptions of the experimental method are very incomplete.

Response: Thanks for the your suggestion. According to advice, we introduced the growth environment and experimental conditions of muskmelon in lines 81-88, described the experimental methods in lines 89-96, and drew a flow chart of the whole growth cycle of muskmelon, as shown in Figure 1.

6. Data processing and analysis. There are no legends in Figure 1 and Figure 2. Then, the selection of pretreatment method, extraction of feature wavelength and comparison of prediction model are all based on only one index, which lacks the reliability of experimental results.

Response: Thanks for the your suggestion. According to advice, in Figure 3, we added the significance diagram of muskmelon growth indicators, and analyzed the differences between different treatments in lines 145-155. Through comprehensive analysis of the differences between muskmelon plant growth indicators and leaf chlorophyll content, we increased the reliability of the experimental results.

7. There is no discussion.

Response: Thanks for the your suggestion. According to advice, we discussed the preprocessing method, VCPA method for feature wavelength extraction and CNN modeling method respectively.

Conclusion. There is no prospect based on the research of this paper in the conclusion, and the logic between paragraphs must be improved.

Response: Thanks for the your suggestion. According to advice, we have rewritten this part and improved the logical relationship between paragraph.

Author Response File: Author Response.doc

Round 2

Reviewer 1 Report

The discussion part is poor. Authors must write the discussion part again. 

Author Response

1.The discussion part is poor. Authors must write the discussion part again.

Response: Thanks for the your suggestion. According to advice, we add the discussion and rewrite the section 4.

The study found that the addition of appropriate red and blue light ratio on the basis of white light in T3 treatment was conducive to the healthy growth of muskmelon, increased chlorophyll content, enhanced photosynthesis, and ultimately improved fruit quality, which was consistent with the results of many studies, such as Chinese cabbage[24], pepper[25]. In Figure 4, the difference exists between T3 treatment and other treatments, mainly because adding red and blue light with appropriate proportion on the basis of white light can promote plant growth, improve chlorophyll content, and is conducive to the growth of muskmelon, which was consistent with the results of Cui et al[26] and Bantis [27].

The correlation coefficient values after Gaussian filtering, MSC and SNV processing have decreased, and the modeling effect has decreased to a certain extent, the minimum value was 0.819, as shown in Table 2. This was because the three preprocessing methods of Gaussian filtering, MSC and SNV can smooth the noise while smoothing the useful information, leading to spectral signal distortion, and reducing the spectral modeling effect after preprocessing[28,29]. After S-G and Detending preprocessing, the correlation coefficient increased and the model performance improved. Convolution smoothing can eliminate noise and minimize the impact of smoothing on useful information, thus optimizing the performance of the model[30]. The Detending algorithm can eliminate the baseline drift of diffuse reflection, thus ensuring the stability and accuracy of the numerical value [31].

In terms of feature wavelength extraction, we found that the model constructed by VCPA method had the best effect by comparing the correlation coefficient and root mean square error of six feature wavelengths extracted based on S-G in Table 3. Because the BMS sampling strategy adopted by VCPA provided the same sampling probability for each variable, the variable space was compressed through EDF to eliminate irrelevant variables, and the MPA idea was adopted to retain the top 10% optimal variable subset, so the probability of important variables being finally selected is relatively high[32]. VCPA feature wavelength extraction method was first confirmed by Yun et al [33]. Now, VCPA method and its combination with other modeling methods were widely used in food quality detection [34], crop protein content [35].

In terms of modeling methods, we established PLSR, LSSVM and CNN models based on VCPA, and concluded that CNN was the best model among the three modeling methods in Table 4. The reason may be that CNN was a deep learning model or a multilayer perceptron similar to an artificial neural network [36]. It was usually used for visual effect image detection, and can learn training features from large drainage matrix data information, and extend its results to the same type of unknown data information. The total amount of parameters required for calculation was greatly reduced, and the reduction of accuracy was effectively prevented. This was confirmed in the research of Yu et al.[37], but the CNN model needs to be debugged for different samples to make the model more generalized.

Author Response File: Author Response.docx

Reviewer 3 Report

The current version of the manuscript is not satisfactory. Although the author has made some changes based on my previous suggestions, most of them are not substantial changes. The author should list the contents and extents in detail of his improvement in the cover letter. It is suggested to give another chance to complete the major revision of this manuscript. And it is still of not unacceptable for publication.

1. Abstract. The author has made some modifications to this part, but the structure of the abstract is not correct, I think it must be rewritten. When describing the experimental steps and research results in the abstract, they should be described separately, rather than directly describing the experimental results obtained after each step of the experiment. And the results of each experiment were only described with words without quantitative data, which weakened its presentation.

2. Introduction. The author has made some modifications to this part, but the suggestions need to be further improved. The first paragraph is too long. It is recommended to reduce it to 2-3 lines and merge it with the second paragraph. In addition, the research status of the third paragraph is suggested to be summarized, and the narrative content is incomplete. This manuscript only describes the current research status of chlorophyll content detection by hyperspectral imaging technology, but does not mention the current status of extraction wavelength characteristics and how to establish chlorophyll prediction model.

3. Materials and Methods. The author has made few adjustments to this part of the content, so it is not completely revised. A brief introduction to the experimental material "Boyang 91" is still not added. It is suggested to add a few sentences of brief introduction to the experimental material in 2.1. In addition, the introduction of experimental methods in 2.3 is only part of 2.3.1, which is still incomplete, so it must be improved.

4. Results and Discussion. The author modified the analysis part of the results to some extent, but not completely. It is recommended that the level 1 heading of this section be modified to avoid placing the results analysis under the same section as the discussion. Second, the legend is still not added in Figure 2, so it is recommended to add. Moreover, the content in Figure 3 is not clear, it is suggested to change the color and enlarge the size. In addition, the content of this chapter is not clearly divided into different levels. For example, the titles of 3.3 and 3.4 are similar, so it is suggested to add three levels of headings if necessary.

5. Discussion. I think the author may have misunderstood my suggestion. In effect, the author needs to create a separate new chapter for the discussion rather than narrating it together with the results analysis in 4, so it is recommended to revise this section again.

6. Conclusion. The author has made few adjustments to this part of the content, so it is not completely revised. This part only describes the conclusion according to the experimental procedure without logic. The different results should be shown to be correlated. Therefore, it is suggested to reorganize it. In addition, this section still does not point out the shortcomings and prospects of the study.

 

Author Response

1. The author has made some modifications to this part, but the structure of the abstract is not correct, I think it must be rewritten. When describing the experimental steps and research results in the abstract, they should be described separately, rather than directly describing the experimental results obtained after each step of the experiment. And the results of each experiment were only described with words without quantitative data, which weakened its presentation.

Response: Thanks for the your suggestion. According to advice, we have rewritten the abstract and have described the trial steps and the experimental results separately, while adding a description of the quantitative data to the test results as follows:

In order to select the light quality suitable for plant growth, a quantitative detection model of chlorophyll content in muskmelon leaves was established to monitor plant growth quickly and accurately. In the paper, muskmelon "Boyang 91" was used as the experimental material, and six different light proportion treatments were set up. Through measuring plant height, stem diameter, number of leaves, nodes and other growth indicators and leaf chlorophyll content, the response difference of muskmelon to different light quality was explored in plant factory. The hyperspectral imaging technology was used to establish a prediction model for the chlorophyll content of muskmelon. The leaves at the fruit bearing stage were taken as the research object to obtain the hyperspectral and chlorophyll content data. The original spectrum was preprocessed by Gaussian Filter, S-G, MSC, SNV, and Detrending, and the correlation coefficient and root mean square error were comprehensively analyzed. Based on the optimal preprocessing method, the characteristic wavelengths were extracted by successive projections algorithm (SPA), competitive adaptive reweighted sampling (CARS), interval variable iterative space shrinking analysis (iVISSA), genetic algorithm and partial least squares (GAPLS), variable combination population analysis (VCPA) and uninformation variable elimination (UVE). By comparing six methods with characteristic wavelength extraction, Partial least-squares regression (PLSR), least squares support vector machine (LSSVM) and convolutional neural network (CNN) were established based on the optimal feature wavelength extraction method. The results showed that there were significant differences between plant height, stem diameter, number of nodes and chlorophyll content of muskmelon treated with T1, T2, T3, T4, T5, CK. Through comprehensive analysis, T3 treatment (Light ratio: 6R/1B/2W, Light quantum flux: 360 μmol/(m2·s), Photoperiod: 12 h) was optimized. The average spectral reflectance data of 216 leaf samples were extracted, and the S-G preprocessing method was selected to preprocess the original spectral data (R_C=0.860, RMSEC=1.806; R_CV=0.790, RMSECV=2.161). By comparing and analyzing the correlation coefficients and root mean square errors of six feature wavelength extraction methods, it was concluded that the VCPA method had the best model effect for feature wavelength extraction (R_P=0.824, RMSEP=1.973). Ten characteristic wavelengths ( 396, 409, 457, 518, 532, 565, 687, 691, 701 and 705 nm) extracted by VCPA method were used to establish the chlorophyll content prediction model, and the chlorophyll content prediction model of S-G-VCPA-CNN had the best performance (Rc=0.9151, RMSEC=1.445; Rp=0.811, RMSEP=2.055). The results of study provide data support and theoretical basis for screening light ratio of other crops, and also present technical support for online monitoring of crop growth in plant factories.

2. The author has made some modifications to this part, but the suggestions need to be further improved. The first paragraph is too long. It is recommended to reduce it to 2-3 lines and merge it with the second paragraph. In addition, the research status of the third paragraph is suggested to be summarized, and the narrative content is incomplete. This manuscript only describes the current research status of chlorophyll content detection by hyperspectral imaging technology, but does not mention the current status of extraction wavelength characteristics and how to establish chlorophyll prediction model.

Response: Thanks for the your suggestion. According to advice, We abbreviated the first paragraph and merge it with the second paragraph, and summarized the research status of the third paragraph, introduced the research status of using hyperspectral imaging technology to detect chlorophyll content, the current status of extracting wavelength features, and how to build a chlorophyll prediction model as follows:

Muskmelon is an important economic and medicinal crop[1] and has a large planting area in Ningxia [2, 3]. The chlorophyll content of crop leaves not only reflects the nutritional status and growth characteristics of plants, but also has a significant impact on crop yield and quality [4]. Therefore, real-time monitoring of chlorophyll content in crop leaves is conducive to accurate monitoring of crop nutrition and growth. Hyperspectral imaging technology was characterized by fast, nondestructive and low cost, enabling direct quantitative analysis of weak spectral differences in vegetation [5]. Many scholars have done a lot of research on non-destructive detection of chlorophyll content in plant leaves by hyperspectral imaging technology [6-8]. The SPA-PLSR model was established by hyperspectral imaging technology for predicting the chlorophyll concentration of peach[9]. Scholars used hyperspectral imaging technology, principal component analysis (PCA) algorithm to select characteristic variables, and PLSR, support vector machine (SVM) and back-propagation neural network (BPNN) algorithms to build hyperspectral estimation models for chlorophyll content in rice leaves, and made comparative analysis. The results showed that the optimal model was PCA-BPNN, and the determination coefficient, root mean square error and relative error of the prediction set were 0.8082, 2.0783 and 4.18% respectively[10]. In the work of Yang Jing et al [11] used hyperspectral imaging technology to establish a BP neural network model to predict the chlorophyll content of rape leaves. The results showed that BP the neural network model was stable and its accuracy has also been greatly improved. Its determination coefficient R2 was between 0.701 and 0.932, and the root mean square error RMSE was between 1.112 and 1.685. The existing research results provided scientific basis for hyperspectral diagnosis of crop chlorophyll content [11-15]. At present, the application of hyperspectral imaging technology for the detection of chlorophyll content of muskmelon leaves was less research.

In this study, we used the muskmelon "Boyang 91" as the experimental material, and picked the leaves of muskmelon fruit stage through different light quality treatment in plant factory. Firstly, we measured the chlorophyll content of muskmelon leaves with chlorophyll instrument, and then collected the hyperspectral images of leaves through the visible near-infrared band hyperspectral imaging system (400~1000 nm). The sample was divided into calibration set (144) and prediction set (72) in a proportion of 2/1, and then the original spectrum was pretreated to select the best pretreatment method. The SPA, CARS, iVISSA, GAPLS, VCPA and UVE were used to extract the characteristic wavelength. The PLSR model of different feature wavelength extraction was compared and analyzed, and the best feature wavelength extraction method was selected. The PLSR, LSSVM and CNN models established under optimized feature wavelength extraction method. Then the best modeling method is selected to predict the chlorophyll content of muskmelon. The quantitative detection model of chlorophyll content in muskmelon leaves was established by combining chemical method with hyperspectral imaging technology to provide technical support for online monitoring of crop growth in plant factories.

3. Materials and Methods. The author has made few adjustments to this part of the content, so it is not completely revised. A brief introduction to the experimental material "Boyang 91" is still not added. It is suggested to add a few sentences of brief introduction to the experimental material in 2.1. In addition, the introduction of experimental methods in 2.3 is only part of 2.3.1, which is still incomplete, so it must be improved.

Response: Thanks for the your suggestion. According to advice, we briefly introduced the melon in 2.1 test materials, and improved the test methods. See 2.3.2, 2.3.3 and 2.3.4.

3. Results and Discussion. The author modified the analysis part of the results to some extent, but not completely. It is recommended that the level 1 heading of this section be modified to avoid placing the results analysis under the same section as the discussion. Second, the legend is still not added in Figure 2, so it is recommended to add. Moreover, the content in Figure 3 is not clear, it is suggested to change the color and enlarge the size. In addition, the content of this chapter is not clearly divided into different levels. For example, the titles of 3.3 and 3.4 are similar, so it is suggested to add three levels of headings if necessary.

Response: Thanks for the your suggestion. According to advice, we added the discussion, see discussion in section 4. At the same time, the titles of 3.3 and 3.4 are merged into the second level heading of 3.3, and the title of 3.3 was changed into the third level headings 3.3.1 and 3.3.2.

Figure 2 is the spectral curve of 216 muskmelon leaves, if adding the legend will make the picture messy. Fig.3 have revise to Fig.4. We changed the color and enlarged the size in Fig.4.

4. I think the author may have misunderstood my suggestion. In effect, the author needs to create a separate new chapter for the discussion rather than narrating it together with the results analysis in 4, so it is recommended to revise this section again.

Response: Thanks for the your suggestion. According to advice, we rewrote the discussion, see discussion in section 4 as follows:

The study found that the addition of appropriate red and blue light ratio on the basis of white light in T3 treatment was conducive to the healthy growth of muskmelon, increased chlorophyll content, enhanced photosynthesis, and ultimately improved fruit quality, which was consistent with the results of many studies, such as Chinese cabbage[24], pepper[25].

In Figure 4, the difference exists between T3 treatment and other treatments, mainly because adding red and blue light with appropriate proportion on the basis of white light can promote plant growth, improve chlorophyll content, and is conducive to the growth of muskmelon, which was consistent with the results of Cui et al[26] and Bantis [27].

The correlation coefficient values after Gaussian filtering, MSC and SNV processing have decreased, and the modeling effect has decreased to a certain extent, the minimum value was 0.819, as shown in Table 2. This was because the three preprocessing methods of Gaussian filtering, MSC and SNV can smooth the noise while smoothing the useful information, leading to spectral signal distortion, and reducing the spectral modeling effect after preprocessing[28, 29]. After S-G and Detrending preprocessing, the correlation coefficient increased and the model performance improved. Convolution smoothing can eliminate noise and minimize the impact of smoothing on useful information, thus optimizing the performance of the model[30]. The Detrending algorithm can eliminate the baseline drift of diffuse reflection, thus ensuring the stability and accuracy of the numerical value [31].

In terms of feature wavelength extraction, we found that the model constructed by VCPA method had the best effect by comparing the correlation coefficient and root mean square error of six feature wavelengths extracted based on S-G in Table 3. Because the BMS sampling strategy adopted by VCPA provided the same sampling probability for each variable, the variable space was compressed through EDF to eliminate irrelevant variables, and the MPA idea was adopted to retain the top 10% optimal variable subset, so the probability of important variables being finally selected is relatively high[32]. VCPA feature wavelength extraction method was first confirmed by Yun et al [33]. Now, VCPA method and its combination with other modeling methods were widely used in food quality detection [34], crop protein content [35].

In terms of modeling methods, we established PLSR, LSSVM and CNN models based on VCPA, and concluded that CNN was the best model among the three modeling methods in Table 4. The reason may be that CNN was a deep learning model or a multilayer perceptron similar to an artificial neural network [36]. It was usually used for visual effect image detection, and can learn training features from large drainage matrix data information, and extend its results to the same type of unknown data information. The total amount of parameters required for calculation was greatly reduced, and the reduction of accuracy was effectively prevented. This was confirmed in the research of Yu et al.[37], but the CNN model needs to be debugged for different samples to make the model more generalized.

6. The author has made few adjustments to this part of the content, so it is not completely revised. This part only describes the conclusion according to the experimental procedure without logic. The different results should be shown to be correlated. Therefore, it is suggested to reorganize it. In addition, this section still does not point out the shortcomings and prospects of the study.

Response: Thanks for the your suggestion. According to advice, we reorganized this part and indicated the shortcomings and prospects in line 280-302. The revised of conclusion is as follows:

In this study, hyperspectral imaging technology was applied to nondestructive detection of chlorophyll content in muskmelon leaves under different light quality treatments in plant factory. T3 treatment (Light ratio: 6R/1B/2W, Light quantum flux: 360 μmol/(m2·s), Photoperiod: 12 h ) had a better performance in plant height, stem diameter, number of nodes and leaf chlorophyll content of muskmelon plant. The prediction model of chlorophyll content in muskmelon leaves was established by fusion spectral. The calibration set and prediction set of 216 muskmelon leaf samples were divided by RS method. The pretreatment method of S-G (R_C=0.860, RMSEC=1.806; R_P=0.790, RMSECV=2.395) was the best among different pretreatment method. The chlorophyll content prediction models were established by characteristic wavelengths extracted of SPA, CARS, iVISSA, GAPLS, VCPA and UVE, and The 10 characteristic wavelengths ( 396, 409, 457, 518, 532, 565, 687, 691, 701 and 705 nm) extracted by VCPA method had the best model effect (R_P=0.824, RMSEP=1.973). Compared with PCR, and LSSVM models, S-G-VCPA-CNN model had the best performance in leaf chlorophyll content (R_C=0.915, R_P=0.811). S-G-VCPA-CNN prediction model of chlorophyll content in muskmelon leaves based on hyperspectral imaging technology provided a reference for rapid detection of other indicators of muskmelon plants and also provided technical support for online monitoring of crop growth in plant factories.

In this paper, the main muskmelon leaves samples were studied and analyzed, and spectral data were collected in an indoor dark box. Although the environmental impact was eliminated, it lacked universality and stability. Many researches can improve the generalization ability and stability of the chlorophyll content prediction model by expanding the number of samples and combining years of data for comprehensive analysis in the field environment at the same time, and the model can be applied to field environment for real-time and rapid nondestructive monitoring. The CNN model structure adopted in this study is fixed and has certain limitations. The model structure can be more optimized. The best model structure can be determined based on the data type and growth period. At present, this research was unable to achieve real-time monitoring of the chlorophyll content of muskmelon leaves. In future, more kinds of algorithms can be explored, and more accurate and stable models can be used to carried out on the development of online monitoring equipment for the chlorophyll content of other plants.

Author Response File: Author Response.docx

Round 3

Reviewer 1 Report

I checked this manuscript. This manuscript can be accepted in its present form.

Author Response

I checked this manuscript. This manuscript can be accepted in its present form.

Response: Thank you very much.

Reviewer 3 Report

According to the review opinions, the author has made some modifications to the manuscript, the quality of the manuscript has been obviously improved. However, there is still much to be modified, I don't think this manuscript has reached the Agronomy-Basel publication criteria.

1. Abstract. The author has made some modifications to this part again, but the suggestions need to be further improved. The "significant difference" mentioned in line 26 of the manuscript does not provide a quantitative description, so it is suggested to supplement. Also, this section has a little more word, so it is recommended that lines 14-26 be further compressed.

2. Introduction. The author has made some modifications to the results, but some parts are not modified, so it is suggested to continue to improve the contents. For example, lines 42-44 lack the logic, and the research status has not been modified. This section recommends writing a summary rather than a simple list. The description content is not complete, so it is recommended to supplement the current research status of extracting wavelength characteristics and establishing chlorophyll prediction model. In addition, the third paragraph of this part does not reflect the highlights of this research, so it is suggested to supplement.

3. Materials and Methods. The author has made only minor adjustments to this part of the content, so it is not effectively revised. The font format in 2.1 is incorrect. Moreover, "2.3" is suggested to be rewritten, because this part is an explanation of the experimental procedures in the manuscript, rather than just a description of which methods were used and what was done in the manuscript. It is suggested that the authors read relevant literature to learn the writing of this part. Also, it could be better understanding to add a flowchart in this section.

4. Results. The author modified the analysis of the results to some extent, but not effectively. The exhibition of figure 4 is not beautiful and rigorous. It is suggested to redesign the color matching and layout. A brief description of Figure 7 and Figure 8 was not given, it is recommended to supplement. Moreover, the starting point, ending point and coordinate interval of the coordinate axes in Figure 8 are inconsistent, which is not convenient for comparison of results. Moreover, the sizes of each subgraph are inconsistent, so it is suggested to modify them.

5. Discussion. The author has made some modifications to this part, but the suggestions must to be further improved. It is suggested to add more comparison contents with other relevant studies, then you should point out the similarities and differences. In addition, it is suggested to further explain and supplement the prediction mechanism of CNN models.

6. Conclusion. The author has made some modifications to the results, but some parts are not modified. You must improve this content. It is recommended that the second paragraph of this section could be relieved to 5-6 lines.

7. Please note the consistency of the font symbols in the images.

Author Response

1.  The author has made some modifications to this part again, but the suggestions need to be further improved. The "significant difference" mentioned in line 26 of the manuscript does not provide a quantitative description, so it is suggested to supplement. Also, this section has a little more word, so it is recommended that lines 14-26 be further compressed.

Response: Thanks for the your suggestion. According to advice,we supplement the quantitative description in lines 18-22, and we compress lines 14-22 as follows:

The hyperspectral imaging technology was used to establish the prediction model for the chlorophyll content of muskmelon. The original spectrum was preprocessed and optimized by different pretreatments. And then the characteristic wavelengths were extracted by six method. Partial least-squares regression (PLSR), least squares support vector machine (LSSVM) and convolutional neural network (CNN) were established for optimal feature wavelength. The results showed that the plant height and stem diameter of T3 treatment were higher than those of other treatments, and their values were 14.48 (cm) and 5.02 (mm) respectively. The chlorophyll content of T3 treatment was the highest and its value was 40.16 (mg/g), which was higher than that of other treatment. Through comprehensive analysis, T3 treatment (Light ratio: 6R/1B/2W, Light quantum flux: 360 μmol/(m2·s), Photoperiod: 12 h) was optimal.

2. The author has made some modifications to the results, but some parts are not modified, so it is suggested to continue to improve the contents. For example, lines 42-44 lack the logic, and the research status has not been modified. This section recommends writing a summary rather than a simple list. The description content is not complete, so it is recommended to supplement the current research status of extracting wavelength characteristics and establishing chlorophyll prediction model. In addition, the third paragraph of this part does not reflect the highlights of this research, so it is suggested to supplement.

Response: Thanks for the your suggestion. According to advice,we add logic to lines 38-42, which complement the current status of research on extracting wavelength features and building chlorophyll prediction models, and supplement the focus of this study in the third paragraph of this section, see lines 55-61.

3. Materials and Methods. The author has made only minor adjustments to this part of the content, so it is not effectively revised. The font format in 2.1 is incorrect. Moreover, "2.3" is suggested to be rewritten, because this part is an explanation of the experimental procedures in the manuscript, rather than just a description of which methods were used and what was done in the manuscript. It is suggested that the authors read relevant literature to learn the writing of this part. Also, it could be better understanding to add a flowchart in this section.

Response: Thanks for the your suggestion. According to advice,we modified the font format in 2.1 and also rewritten 2.3.

4. The author modified the analysis of the results to some extent, but not effectively. The exhibition of figure 4 is not beautiful and rigorous. It is suggested to redesign the color matching and layout. A brief description of Figure 7 and Figure 8 was not given, it is recommended to supplement. Moreover, the starting point, ending point and coordinate interval of the coordinate axes in Figure 8 are inconsistent, which is not convenient for comparison of results. Moreover, the sizes of each subgraph are inconsistent, so it is suggested to modify them.

Response: Thanks for the your suggestion. According to advice, we modified the color matching and layout of Figure 4, briefly describe Figure 7 in line 248-254, and Figure 8 in line 265-270 , and also adjusted the beginning, end point and coordinate interval of the axes in Figure 8, and adjusted the size of each figure in the text.

5. The author has made some modifications to this part, but the suggestions must to be further improved. It is suggested to add more comparison contents with other relevant studies, then you should point out the similarities and differences. In addition, it is suggested to further explain and supplement the prediction mechanism of CNN models.

Response: Thanks for the your suggestion. According to advice,we add a comparison with other relevant studies and then point out similarities and differences. Meanwhile, the last paragraph of this section adds to the prediction mechanism of the CNN model.

6. The author has made some modifications to the results, but some parts are not modified. You must improve this content. It is recommended that the second paragraph of this section could be relieved to 5-6 lines.

Response: Thanks for the your suggestion. According to advice, we have made a reduction in the second paragraph of this section.

7. Please note the consistency of the font symbols in the images.

Response: Thanks for the your suggestion. According to advice, we modified the font size in the image and ensured that the size and font symbol of each image were consistent.

Author Response File: Author Response.doc