Maize Nitrogen Grading Estimation Method Based on UAV Images and an Improved Shufflenet Network

Round 1

Reviewer 1 Report

This paper aimed to explore the utilization of machine learning and deep learning techniques for the classification and measurement of nitrogen levels in maize plots using an Unmanned Aerial Vehicle (UAV), and below are my comments that may help the authors further improve their manuscript:

1. Please revise the author’s affiliation numbers with the subscribed numbers for the authors.

2. Figures and Tables numbers are not cited throughout the manuscript’s text; please cite these within the text instead of Error! Reference source not found.

3. Line 140: The authors should add the geographical coordinates of the experimental maize field; please add the longitude, latitude, and altitude.

4. In Table 1, please correct this Normalized read index to be Normalized red index.

5. Line 253: Please add the full definition of Support vector machine (SVM) at the beginning of the sentence.

6. Please revise and correct line 372.

7. In Figure 14, the abbreviation of SVM is written in the figure’s caption as SVC; please revise and amend all abbreviations.

 

8. What about the limitations of this study? The authors should add the limitations of this study throughout the manuscript’s text.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

see the attachment 

Comments for author File: Comments.pdf

Minor editing of English language required

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The methods proposed based on most recent advances in the sectors of machine learning and deep learning approaches, happen enough accurate and effective to be advantageously used in the areas of crop management support through precision agriculture innovative approaches. However, in my opinion, some aspects of the implemented methodologies, should be better clarified, explained and deepened. In general, the vegetation reflectance in the NIR range, which is sensibly higher than  that of visible wavelengths, plays a predominant role in the vegetation monitoring. The sensibly lower R and B components of the visible solar radiation, is absorbed for photosynthesis while the G one is reflected. Here only the RGB components of UAV camera were exploited while other spectral components were introduced for spectral feature assessment without explain their derivation that must be reported.  From this point of view a more effective synthesis formula based on the above cited  physical aspect should  have negative coefficient for absorbed components and positive for the green one.

 

Thus some aspects about the methodology in situ data and preprocessing should be better explained and supported.  In particular the following points should be enhanced and deepened:

 

·         usually most of biophysical parameters of vegetation were extensively estimated using spectral vegetation indices based on the NIR reflectance while the estimates based only on the RGB components used in this context happen less reliable and robust and this should be more suitably supported;

·         the in situ data acquisition methods description should be better explained and clarified, starting from description of the area of interest features like topography. These aspects may have radiometric effects on the acquired images depending on the sun height and acquisition geometry. These effects and their contributions in terms of BRDF anisotropy on the reported results taking into account the sun height (day time) acquisition and sensor geometry (angle correction 31°), should to be better discussed and deepened, considering also the significant 3d structure (relevant height) of the maize plants;

·         the quantitative classes representation in term of nitrogen content % (PNC) should be provided.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The authors have addressed my comments point-by-point, and I am satisfied with their responses.

Reviewer 2 Report

author has responded and incorporated all the comments asked 

Minor editing of English language required