In Situ Nondestructive Detection of Nitrogen Content in Soybean Leaves Based on Hyperspectral Imaging Technology
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors
The manuscript entitled „In-situ nondestructive detection of nitrogen content in soybean leaves based on hyperspectral imaging technology” presents interesting study on application of remote sensing methods for evaluation of nitrogen content in plants. The study is interesting and manuscript is quite well prepared, however contains some drawbacks.
Please adjust formatting of the manuscript to the guidelines for authors. Title should be written using capital letters.
Line 126: In the manuscript resolution (number of pixels) is presented but what was the pixel size from such distance from the measurements were performed?
Line 126: “the spectral resolution 0.745 nm” but Line 175: “with the resolution of 0.65 nm”. Why the spectral resolution is different?
Line 110: “Each soybean variety was set up with 15 fertilizer treatments”. Could you explain what range of the nitrogen doses was applied. Moreover, because soybean can uptake nitrogen from the air through symbiosis with bacteria, has such symbiosis been observed?
I suggest to present basic information about the treatments, ie. nitrogen doses which were applied together with mean content of nitrogen in leaves for each treatment.
Could you explain what does it mean m and p in equation for RMESC? It is RMESC or RMSEC? There are two abbreviations for the same parameter.
Table 2: In what units are presented values of RMSEC. The same question is for Table 3.
Fig. 10: In what units is presented nitrogen content? Please notice that nitrogen content in leaves is in very large range. In such case strength of the relationships expressed as coefficient of determination or other similar relative parameter is very strong but error of estimation can be very high.
Fig. 11: Please add information if the nitrogen content is estimated or measured. Please be more specific.
Line 502-509: The sections are not completed.
Please adjust formatting of the references to the guidelines for authors of the journal.
Author Response
Thank you for your review.
Colour code: comments by reviewers are in red texts.
Reviewer #10
Point 1: Please adjust formatting of the manuscript to the guidelines for authors. Title should be written using capital letters.
Response 1: Revision has been made, thanks.
Point 2: In the manuscript resolution (number of pixels) is presented but what was the pixel size from such distance from the measurements were performed?
Response 2: Thanks, in this paper, the hyperspectral camera has a spectral scanning range of 400~1000 nm, a spectral resolution of 2.8 nm, a spectral interval of 0.65 nm, and a pixel size of 1344 (rows) × 1024 (columns). During the experiment, the distance between the object and the camera was 60 cm, and the actual size of the object represented by one pixel in the image was 0.16 nm.
Point 3: Line 126: “the spectral resolution 0.745 nm” but Line 175: “with the resolution of 0.65 nm”. Why the spectral resolution is different?
Response 3: We are very sorry for the error. It has been rephrased as follows, thanks. The spectral resolution is 2.8 nm and the spectral interval is 0.65 nm.
Point 4: “Each soybean variety was set up with 15 fertilizer treatments”. Could you explain what range of the nitrogen doses was applied. Moreover, because soybean can uptake nitrogen from the air through symbiosis with bacteria, has such symbiosis been observed? I suggest to present basic information about the treatments, i.e., nitrogen doses which were applied together with mean content of nitrogen in leaves for each treatment.
Response 4: Thanks very much for the suggestions. Nitrogen and mean content of nitrogen in leaves for each treatment have been added. Moreover, the focus of this study is not symbiosis. The statement is rephrased as follows.
It is well known that there is a symbiotic relationship between soybean and rhizobium, and the nitrogen fixation of rhizobium will provide a source of nitrogen for the growth of soybean. However, our research focuses on the establishment of a rapid detection model of soybean leaf nitrogen content and the realization of in-situ non-destructive detection of nitrogen content of soybean leaves, while the establishment of a rapid detection model of soybean leaf nitrogen content depends on the establishment of a larger number of nitrogen content data of soybean leaves with a certain concentration gradient. Therefore, in this study, fertilization treatment was used to obtain the samples of nitrogen content of soybean leaves with a certain concentration gradient, in order to establish a strong adaptability of the nitrogen content detection model of soybean leaves.
Line: 96-100
Three-factor quadratic orthogonal regression was used to quantitatively fertilize soybean to obtain nitrogen gradient, in order to obtain soybean leaf samples with a certain concentration of nitrogen gradient. Urea was applied as nitrogen fertilizer in the experiment, and the range of nitrogen application was 0 ~ 70 kg/hm2. The specific fertilization scheme was shown in Table 1.
Table 1 Design table of quadratic orthogonal regression experiment for soybean fertilization.
|
Processing number |
z1(N) |
z2(P) |
z3(K) |
N（kg/hm2） |
P（kg/hm2） |
K（kg/hm2） |
|
1 |
1 |
1 |
1 |
63.80 |
100.25 |
127.59 |
|
2 |
1 |
1 |
-1 |
63.80 |
100.25 |
12.41 |
|
3 |
1 |
-1 |
1 |
63.80 |
9.75 |
127.59 |
|
4 |
1 |
-1 |
-1 |
63.80 |
9.75 |
12.41 |
|
5 |
-1 |
1 |
1 |
6.20 |
100.25 |
127.59 |
|
6 |
-1 |
1 |
-1 |
6.20 |
100.25 |
12.41 |
|
7 |
-1 |
-1 |
1 |
6.20 |
9.75 |
127.59 |
|
8 |
-1 |
-1 |
-1 |
6.20 |
9.75 |
12.41 |
|
9 |
1.21541 |
0 |
0 |
70.00 |
55.00 |
70.00 |
|
10 |
-1.21541 |
0 |
0 |
0.00 |
55.00 |
70.00 |
|
11 |
0 |
1.21541 |
0 |
35.00 |
110.00 |
70.00 |
|
12 |
0 |
-1.21541 |
0 |
35.00 |
0.00 |
70.00 |
|
13 |
0 |
0 |
1.21541 |
35.00 |
55.00 |
150.00 |
|
14 |
0 |
0 |
-1.21541 |
35.00 |
55.00 |
0.00 |
|
15 |
0 |
0 |
0 |
35.00 |
55.00 |
70.00 |
Note: z1 was the level of nitrogen factor; z2 was the phosphorus factor level; z3 was the potassium factor level.
Point 5: Could you explain what does it mean m and p in equation for RMESC? It is RMESC or RMSEC? There are two abbreviations for the same parameter.
Response 5: Thanks very much for the suggestions. I am sorry that this part was not clear in the original manuscript. I should have explained that m was the number of samples in the calibration set, n was the number of samples in the prediction set, and p was the best principal component in the modeling process (also known as the number of latent variables). RMSEC represent the root mean square error of the calibration set, and RMSEP represent the root mean square error of the prediction set.
Point 6: Table 2: In what units are presented values of RMSEC. The same question is for Table 3.
Response 6: The unit representing the value of RMSEC is mg/g. Revision has been made, thanks.
Point 7: Fig. 10: In what units is presented nitrogen content? Please notice that nitrogen content in leaves is in very large range. In such case strength of the relationships expressed as coefficient of determination or other similar relative parameter is very strong but error of estimation can be very high.
Response 7: Thanks very much for the suggestions. I am sorry that this part was not clear in the original manuscript. The statement is rephrased as follows.
Revision has been made, thanks. As stated by the reviewer, the range of nitrogen content in leaves is very large. In such case, the strength of the relationships expressed as coefficient of determination or other similar relative parameter is very strong, but the error of estimation can be very high. Therefore, we used the RPD to evaluate the established model to improve the accuracy. Generally, when the RPD value was less than 1.5, it indicated that the performance of built model was poor and cannot be used for predictive analysis. When the RPD value was between 1.5 and 2.0, it means that the prediction effect of the established prediction model was general, and the sample can be roughly estimated. When RPD value was lager than 2.0, it shows that the model has good predictive ability.
Fig. 10:
Point 8: Fig. 11: Please add information if the nitrogen content is estimated or measured. Please be more specific.
Response 8: Revision has been made, thanks.
Fig. 11:
Point 9: Line 502-509: The sections are not completed.
Response 9: Revision has been made, thanks.
Point 10: Please adjust formatting of the references to the guidelines for authors of the journal.
Response 10: Revision has been made, thanks.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for Authors
20/03/2024
Dear authors,
In the manuscript In-situ nondestructive detection of nitrogen content in soybean leaves based on hyperspectral imaging technology you combined hyperspectral imaging technology with chemometrics methods and used to detect the nitrogen content of soybean leaves, to achieve the rapid, non-destructive and in-situ detection of nitrogen content in soybean leaves. Soybean leaves under different fertilization treatments were used as the research object, and the hyperspectral imaging data and the corresponding nitrogen content data of soybean leaves at different growth stages were obtained.
General comments
The study is interesting and experimental results indicate the possibility of wider use of hyperspectral imaging technology for soybean nutrition diagnosis, growth evaluation and fertilization management in frame of precise agronomy.
Abstract is too long, twice as long as recommended (yours is 401, and should be up to 200 words). In the Introduction, you should briefly state the motivation for solving the problem (not in detail), state the problem and how you solved it, but only the main result.
The biggest problem of the Introduction is the lack of commenting on existing methodologies, references for it and highlighting what you bring new to this field. Also missing is an announcement of what follows in the manuscript.
It is not clear from the Materials and Methods chapter how and why you determined the methods of using fertilization. You need to explain how you determined the parameters for carrying out the measurements and controlling the results.
In Discussion you should discuss your results and why they are like that. Instead, you list only the results and write a text that is specific to the Introduction, which others have done before.
A lot of time and effort was invested in this work, but it is not clear what was intended to be achieved with the results. The authors do not introduce new methods or parameters, they only conduct analyzes of existing methods (statistical, mostly). This work is more professional and not scientific.
Specific comments (are in the manuscript)
- Lines 23-32 – This text is too detailed for the Abstract.
- Line 44 - Typographical error, delete space.
- Lines 61-62 - Here you are making claims that you have not backed up with references.
- Lines 71-72 - Here you are making claims that you have not backed up with references.
- Lines 87-99 - Here you are making claims without any references, again.
- Lines 106-109 - This sentence should be clarified. What did you mean to say with it?
- Line 118 - Typographical error, the letter 'n' is missing at the end of the word.
- Lines 430-441 - This text does not belong in Discussion, it belongs in Introduction.
- Line 441 - Typographical error, lowercase letter.
Best regards
Comments for author File: Comments.pdf
Author Response
Thank you for your review.
Colour code: comments by reviewers are in red texts.
Reviewer #14
Point 1: Abstract is too long, twice as long as recommended (yours is 401, and should be up to 200 words). In the Introduction, you should briefly state the motivation for solving the problem (not in detail), state the problem and how you solved it, but only the main result.
Response 1: Thank you very much for the valuable suggestions. It has been rephrased as follows.
Line: 13-34
Abstract: In this paper, hyperspectral imaging technology combined with chemometrics methods were used to detect the nitrogen content of soybean leaves, to achieve the rapid, non-destructive and in-situ detection of nitrogen content in soybean leaves. Soybean leaves under different fertilization treatments were used as the research object, and the hyperspectral imaging data and the corresponding nitrogen content data of soybean leaves at different growth stages were obtained. Seven spectral preprocessing methods, such as Savitzky-Golay smoothing (SG), first derivative (1-Der), and direct orthogonal signal correction (DOSC), were used to establish the quantitative prediction models for soybean leaf nitrogen content, and the quantitative prediction models of different spectral preprocessing methods for soybean leaf nitrogen content were analyzed and compared. On this basis, successive projections algorithm (SPA), genetic algorithm (GA) and random frog (RF) were employed to select the characteristic wavelengths, and compress the spectral data. The results showed that: (1) The full-spectrum prediction model of soybean leaf nitrogen content based on DOSC pretreatment was the best. (2) The PLS model of soybean leaf nitrogen content based on the five characteristic wavelengths had the best prediction performance, with Rp2 of 0.9466. (3) The spatial distribution map of soybean leaf nitrogen content was generated in pixel manner using the extracted 5 characteristic wavelengths and DOSC-RF-PLS model. The nitrogen content level of soybean leaves can be quantified in a simple way, this provides a foundation for rapid in situ non-destructive detection and spatial distribution difference detection of soybean leaf nitrogen. (4) The overall results illustrated that hyperspectral imaging technology was a powerful tool for spatial prediction of nitrogen content in soybean leaves, which provided a new method for the spatial distribution of soybean nutrient status and the dynamic monitoring of growth status.
Point 2: The biggest problem of the Introduction is the lack of commenting on existing methodologies, references for it and highlighting what you bring new to this field. Also missing is an announcement of what follows in the manuscript.
Response 2: Thank you very much for the professional comments and valuable suggestions. It has been rephrased as follows.
Line: 36-83
Soybean (Glycine max (L.) Merrill), is the important source of plant protein and vegetable oi for human beings. It have a wide range of uses worldwide, such as for food, vegetable oil, and feed [1]. There is a huge gap between China’s soybean supply and consumption demand. Therefore, it is an important measure to solve the contradiction between supply and demand of soybean in China and ensure the safety of soybean industry and national grain and oil security by improving the yield and quality of soybean in China and increasing the proportion of soybean cultivation [2].
Nitrogen is an important part of protein, nucleic acid, phospholipid, chlorophyll and some hormones in soybean plant, and it is also one of the important factors limiting soybean growth and high yield. The rapid and effective monitoring of nitrogen content during soybean growth is an important prerequisites and foundations for guiding soybean fertilization and achieving high quality and high yield of soybean. Chemical methods, such as Kjeldahl nitrogen determination, indophenol blue colorimetric, and Dumas combustion, are commonly used for nitrogen detection. These methods have problems of long detection cycle, complex operation, and destructiveness, which make the continuous determination of nitrogen in time and space impossible[3].
Hyperspectral imaging technology is an organic combination of spatial imaging technology and spectral technology. The image information and spectral information of the research object could be obtained at the same time using hyperspectral imaging technology. So it is an effective tool to study the internal information content and spatial distribution of the research object[4,5]. The hyperspectral imaging technology has been successfully applied in remote sensing, food, agriculture, microbiology and pharmaceutical fields[6,7]. In addition, the detection of crop nitrogen content based on hyperspectral imaging technology has achieved good results in wheat, corn, rapeseed, citrus and other crops. Hyperspectral imaging technology has been applied to measure the nitrogen content in wheat leaves, and the regional spatial distribution map of nitrogen content in wheat leaves at flagging and flowering stages was generated[8]. Goel et al.analyzed the hyperspectral images of maize canopy under nitrogen stress and weed stress, and found that the spectral reflectance at 498 nm and 671 nm could effectively reflect the difference of nitrogen level in maize[9]. By using the spectrum-graph unity feature of hyperspectral imaging technology, differences in nitrogen content within the same leaf or among different leaves can be analyzed intuitively and effectively. However, to the best of our knowledge, few studies have been conducted to study the distribution of nitrogen content in soybean leaves based on hyperspectral imaging technology, the relationship between the spectral reflectance of soybean leaves and the leaf nitrogen content is not clear, and it remains to be demonstrated whether the use of hyperspectral imaging technology can realize the effective detection and continuous monitoring of the nitrogen content of soybean leaves.
Therefore, in this study, the relationship between the spectra of soybean leaves and nitrogen content under different fertilization treatments was analyzed, the effects of spectral pre-processing and characteristic wavelength selection methods on the prediction model of nitrogen content of soybean leaves were investigated, and the optimal detection model for the detection of nitrogen content of soybean leaves was optimized. On this basis, the spatial distribution map of soybean leaf nitrogen content was generated to provide a new method for the spatial prediction of soybean nutrient status and the dynamic monitoring of the growth status, which in turn provided a basis for the application of fertilizer decisions during soybean growth.
Point 3: It is not clear from the Materials and Methods chapter how and why you determined the methods of using fertilization. You need to explain how you determined the parameters for carrying out the measurements and controlling the results.
Response 3: Thanks very much for the suggestions. I am sorry that this part was not clear in the original manuscript. It has been rephrased as follows.
Line: 96-100
Three-factor quadratic orthogonal regression was used to quantitatively fertilize soybean to obtain nitrogen gradient, in order to obtain soybean leaf samples with a certain concentration of nitrogen gradient. Urea was applied as nitrogen fertilizer in the experiment, and the range of nitrogen application was 0 ~ 70 kg/hm2. The specific fertilization scheme was shown in Table 1.
Table 1 Design table of quadratic orthogonal regression experiment for soybean fertilization.
|
Processing number |
z1(N) |
z2(P) |
z3(K) |
N（kg/hm2） |
P（kg/hm2） |
K（kg/hm2） |
|
1 |
1 |
1 |
1 |
63.80 |
100.25 |
127.59 |
|
2 |
1 |
1 |
-1 |
63.80 |
100.25 |
12.41 |
|
3 |
1 |
-1 |
1 |
63.80 |
9.75 |
127.59 |
|
4 |
1 |
-1 |
-1 |
63.80 |
9.75 |
12.41 |
|
5 |
-1 |
1 |
1 |
6.20 |
100.25 |
127.59 |
|
6 |
-1 |
1 |
-1 |
6.20 |
100.25 |
12.41 |
|
7 |
-1 |
-1 |
1 |
6.20 |
9.75 |
127.59 |
|
8 |
-1 |
-1 |
-1 |
6.20 |
9.75 |
12.41 |
|
9 |
1.21541 |
0 |
0 |
70.00 |
55.00 |
70.00 |
|
10 |
-1.21541 |
0 |
0 |
0.00 |
55.00 |
70.00 |
|
11 |
0 |
1.21541 |
0 |
35.00 |
110.00 |
70.00 |
|
12 |
0 |
-1.21541 |
0 |
35.00 |
0.00 |
70.00 |
|
13 |
0 |
0 |
1.21541 |
35.00 |
55.00 |
150.00 |
|
14 |
0 |
0 |
-1.21541 |
35.00 |
55.00 |
0.00 |
|
15 |
0 |
0 |
0 |
35.00 |
55.00 |
70.00 |
Note: z1 was nitrogen factor level, z2 was the phosphorus factor level; z3 was the potassium factor level.
Point 4: In Discussion you should discuss your results and why they are like that. Instead, you list only the results and write a text that is specific to the Introduction, which others have done before.
Response 4: Thank you very much for the professional comments. It has been rephrased as follows.
Line: 431-470
In the detection of plant nutrient information, traditional crop nutrient diagnostic methods have some shortcomings in the detection of nutrient information. For instance, the appearance diagnosis method is subjective, easy to misdiagnose, can not realize the effect of active prevention[23,24]. The chemical diagnosis method has poor timeliness, complex operation, and high destructiveness [25,26].The chlorophyll meter diagnosis method has a small detection range, high environmental requirements, and can not accurately detect the nutrient status of a larger area[27]. Therefore, the traditional nutrient diagnostic methods have great limitations and latency in practical applications, and cannot adapt to the practical needs of precision agriculture for rapid, real-time, nondestructive, and large-area detection of crop nutrient information. In this paper, we proposed a method for detecting the nitrogen content of soybean leaves based on hyperspectral imaging technology, with a high detection accuracy of Rp2 of 0.9466. The nitrogen content of soybean leaves can be detected in a rapid, real-time and non-destructive manner, which could provide scientific guidance for the monitoring of nutrient dynamics and the reasonable regulation of fertilizers in the process of soybean production.
In the aspect of model accuracy, Kang Kai et al. Used multispectral techniques to detect the nitrogen content of canopy leaves in soybean fields, and the coefficient of determination R² of the established regression model ranged from 0.8 to 0.9 [28]. Wang Lifeng et al. developed a 1st-SPA-PLS model for monitoring nitrogen content of corn leaves in field using hyperspectral imaging technology, with an accuracy of Rp2 of 0.749 [29]. Qin Zhanfei et al. estimated the total nitrogen content of rice leaves in the yellow diversion irrigation area using hyperspectral imaging technology, and the modeling accuracy R² was 0.673 [30]. The accuracy of the soybean leaf nitrogen content detection model established in this study was slightly higher than that of the nitrogen content detection models established in other studies. On the one hand, it is probably that the greenhouse potting method was used for fertilizer control in this study, and the control of nitrogen fertilizer gradient is relatively accurate and stable. On the other hand, it may be that indoor hyperspectral imaging data acquisition was stable compared with the outdoor field measurement environment, with relatively small interfering factors and noise signals, therefore the stability and reliability of the established nitrogen content detection model were higher. In addition, most of the research data collection for nitrogen content detection focused on one or fewer fertility stages , due to the great differences in nutrient composition of the crop at different growth stages, the experimental samples obtained at one or more fertility stages had a small nitrogen gradient, whereas the nitrogen detection in this study was carried out throughout the whole fertility stage of soybean, and it was possible to obtain samples of soybean leaves with large nitrogen gradients, which may be another important reason for the slightly higher accuracy of the nitrogen content detection model established in this study than that of the others.
Point 5: A lot of time and effort was invested in this work, but it is not clear what was intended to be achieved with the results. The authors do not introduce new methods or parameters, they only conduct analyzes of existing methods (statistical, mostly). This work is more professional and not scientific.
Response 5: Thanks very much for the suggestions. Due to the complexity and uncertainty of crop growth, the focus of this study was to explore the feasibility of in situ, non-destructive and rapid detection of nitrogen content in soybean leaves using hyperspectral imaging technology. Therefore, by analyzing the relationship between the spectral reflectance of soybean leaves and the nitrogen content of leaves, the effects of spectral pre-processing and characteristic wavelength selection methods were studied on the prediction model of nitrogen content of soybean leaves, the optimal detection model for the detection of nitrogen content of soybean leaves was optimized, and the DOSC-RF-PLS was determined as the optimal detection model for the nitrogen content of soybean leaves, with Rp2 of 0.9466. On the basis, the spatial distribution map of nitrogen content in soybean leaves was generated and visualized, which can provide scientific guidance for the monitoring of the nutrient dynamics and the rational regulation of fertilizers in the process of soybean planting.
Point 6: Lines 23-32 – This text is too detailed for the Abstract.
Response 6: Revision has been made, thanks.
Point 7: Line 44 - Typographical error, delete space.
Response 7: Sorry for the mistake. Revision has been made, thanks.
Point 8: Lines 61-62 - Here you are making claims that you have not backed up with references.
Response 8: Thanks, the references are added.
Point 9: Lines 71-72 - Here you are making claims that you have not backed up with references.
Response 9: Thanks, the references are added.
Point 10: Lines 87-99 - Here you are making claims without any references, again.
Response 10: Thanks, the references are added.
Point 11: Lines 106-109 - This sentence should be clarified. What did you mean to say with it?
Response 11: Thanks very much for the suggestions. I am sorry that this part was not clear in the original manuscript. This sentence means the experimental site and the climatic conditions for the implementation of the experiment.
Line: 88-94
The experiment was conducted from March to August 2017 in experimental base of National Engineering Research Centre of Intelligent Equipment for Agriculture, Xiaotangshan Town, Changping District, Beijing, China (116°44’E, 40°18’N). This area has a warm-temperate continental semi-humid and semi-arid monsoon climate, with an average annual temperature of 11.8°C, an average annual frost-free period of 203 days, an average annual sunshine hour of 2816 hours, and an average annual precipitation of 584 mm.
Point 12: Line 118 - Typographical error, the letter 'n' is missing at the end of the word.
Response 12: Sorry for the mistake. Revision has been made, thanks.
Point 13: Lines 430-441 - This text does not belong in Discussion, it belongs in Introduction.
Response 13: Revision has been made, thanks.
Point 14: Line 441 - Typographical error, lowercase letter.
Response 14: Revision has been made, thanks.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for Authors
The manuscript was improved according all my suggestions but the response for one questions is not sufficient. Please see this question more carefully because your reply is not sufficient:
Could you explain what does it mean m and p in equation for RMESC? It is RMESC or RMSEC? There are two abbreviations for the same parameter.
Author response
Point 1: Could you explain what does it mean m and p in equation for RMESC? It is RMESC or RMSEC? There are two abbreviations for the same parameter.
Response 1:
We are sorry for the mistake of the formula. After consulting the relevant references, we have found the problem and corrected the formula. The specific modifications are as follows :
We have changed from “” to “”, and have changed from “” to “”. The m in equation for RMESC was the number of samples in the calibration set, and the n in equation for RMESP was the number of samples in the prediction set.
The RMSEC represent the root mean square error of the calibration set and RMSEP represent the root mean square error of the prediction. The two abbreviations could be found in lines 189-190. RMESC and RMESP were misspellings of RESEC and RMSEP respectively. We sincerely apologize for this spelling error and thank the reviewers for their conscientiousness.
Reviewer 2 Report
Comments and Suggestions for Authors
29/03/2024
Dear authors,
In the manuscript In-situ nondestructive detection of nitrogen content in soybean leaves based on hyperspectral imaging technology you combined hyperspectral imaging technology with chemometrics methods and used to detect the nitrogen content of soybean leaves, to achieve the rapid, non-destructive and in-situ detection of nitrogen content in soybean leaves. Soybean leaves under different fertilization treatments were used as the research object, and the hyperspectral imaging data and the corresponding nitrogen content data of soybean leaves at different growth stages were obtained.
General comments
The study is interesting and experimental results indicate the possibility of wider use of hyperspectral imaging technology for soybean nutrition diagnosis, growth evaluation and fertilization management in frame of precise agronomy. You’ve looked back at all the comments I gave in the first version, and now your handwriting looks more readable to me and better highlights what you wanted to show. Thank you. I hope you are more satisfied with this version than the previous one.
Specific comments
I have no specific comments
Best regards
