A New Method for Extracting Individual Plant Bio-Characteristics from High-Resolution Digital Images

Round 1

Reviewer 1 Report

This is an interesting study. A study area map and other details (e.g., climate, vegetation, precipitation, population etc.) of the study site would be very important for the readers.

Certainly I would recommend adding a study area map and possible details of  the study are rather than name and location/latitude and longitude of the study.  

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The authors introduce a new method for extracting crop leave area, center point, and NDVI at high spatial resolution with acceptable reliability. The research design is reasonable, and the manuscript is well-organized. Yet, I have some concerns and questions for the authors to consider and address. And I hope this review is constructive.

 

Title:

1. I understand why you include the term “in row trails” here but I don’t think it’s necessary to do so. Also, I don’t think you’ve obtained the real biomass information (e.g., NPP), instead you extract some basic plant bio-characteristics. I recommend keeping the title straightforward and attractive like “A new method for extracting individual plant bio-characteristics from high-resolution digital images”.

Abstract:

Line 21: Do you really think the current correlation is “good”? How about thinking of another word and make the sentence sounds modest?

Keywords:

I don’t think adding the “correlation coefficient” to the list is a good idea.

Problem statements:

1. Line 86: I notice you use the term “trial data” quite often. Do you mean experimental data from fieldwork? Then how about “field data”? I think the latter one is more common.
2. Line 99: Please illustrate more about the GoPro image properties, such as its spatial resolution, band information. You’re submitting your work to a journal in the Remote Sensing field. So please talk about the images you obtained.
3. Line 102 and Figure 1(b): Did you make 1(b) using 30-m Landsat? I doubt because I don’t think the coarse 30-m spatial resolution can capture the details like space between individual plants in 1(b).
4. Line 115-119: I suggest reconstructing this sentence. It’s too long and took the audience extra time to digest.

Methods:

1. Line 218: Why you set up the coefficient as “0.6” in equation (2)? Also, please explain more on the upper and lower limits of the range of i. Why 0.2r and 0.8r?
2. Line 269 and 272: Why you minus 1 in equations (4) and (5)?
3. Line 278: Where is equation (10) and (11)?

Results and Discussions:

1. Line 345: The biomass change range is really large. I’m curious about 428g, does it mean that the change of fresh biomass weight can reach up to 428g in different seasons?
2. Figure 13: The label of your x-axis is wrong, I guess. It should be “DOY”, not plant No. And also can you add both the y-axis of “Normalized Fresh Weights” (you already have done) and the y-axis of “Normalized Area” to Figure 13? One on the right and another on one the left.
3. Line 390: Or maybe when you apply your method to another indicator, you could see more significant improvements? You can add this to the discussion section.

Conclusions:

1. Line 396: “:” rather than “;”.

And I would like to encourage the authors to improve the quality of their Plot figures such as Figure 12, 13. You can use Matlab or R or python to generate more professional product for publication.

Author Response

Please see the attachment.

Reviewer 3 Report

1. The goal is to automatically extract phenomic traits such as area and NDVI value of each plant from each field trial’s TIFF file image. However, phenomic traits consist of many factors. Why this paper only adopted NDVI value should be addressed.
2. Whether a plant bounding box contained a plant was determined by identifying the centre point of a plant. However, it is hard to define plant bounding boxes in Figure 6. (d), a background corrected image, unlike Figure 7 whose distribution of greenness or NDVI values is so obvious to identify an individual plant. How to find a maximum and a minimum without a clear plant’s edges?
3. Figure 10. (d) seems to extract over-size circular plant regions for most plants. Is a too large radii adopted? Could this problem be modified?
4. The correlation coefficients between the area of circular plant regions and fresh weights for four field trial images from four timepoints are shown in Figs. 12. Shall the authors expect to predict the fresh weight weights based on the area of circular plant regions? If so, axis X and axis Y should be exchanged.
5. Bounding box and bound box should be consistent.
6. The bounding boxes have intensity values satisfying ??(?,?) ≥ ?.?∗????. Why is 0.6? Each plant-box is constrained to not include the top and bottom 20% of the rows. Why is 20%?
7. Results and Discussions and 5. Conclusions are suggested to further extending.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 4 Report

Dear authors,

Thanks for your paper. I think, that you describe methods for of image processing, but I don’t see there real benefit of this methods and real usability. Please if you will resubmit proposal describe in introduction clearly purpose of this analysis. What will be benefit, and how this analysis can be used to understand better phenology of crop.  This topic is missing for me.

Other things, which I didn’t see clearly explained in paper is scale/resolution of images. I think, that this information is key for the possibilities to understand reusability of this approach.

With this is also related question, how this method can be used. Can such solution to be implemented of board computer as some method of edge computing.  Or will it be used in links with drones.

I don’t see possibilities, how such solution can be used for interpretation of satellite images or additional analysis of phenology.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

I can tell the authors spend time and effort to answer my questions and make improvements accordingly. I think the current version is good.

Author Response

Thank you for your valuable time and useful suggestions to review the article.

Reviewer 3 Report

No more comments.

Author Response

Thank you for your valuable time and useful suggestions on the article.

Reviewer 4 Report

Dear authors,

thanks, paper is now more clear, but seems to me that all bio characteristic are described by area and NDVI. Can you please explain, how this information can improve genomic knowledge and its relation to phenotypes. This part is missing in paper. Then it is not fully evident benefit for plan breading.

Algorithms and methods are clear, results are described, but I have problem to find real benefit of this experiments.

Author Response

Thank you for these comments.  We have added text in the introduction providing more detail how these phenotypes are useful for breeding.  Please let us know if you would like additional detail.

Phenomic bio-characteristics such as NDVI or plant area can be correlated or predictive of plant biomass yield, which is the main production phenotype in forage species and is a characteristic contributing to grain yield in other crops [21, 22]. Bio-characteristics, if sufficiently correlated, can then be used as proxy phenotypes for biomass in genomic se-lection to select the best populations and generate genetic gain over generations.  Furthermore, as image derived bio-characteristics are non-destructive, they can be collected at multiple time-points during the growth cycle of crops, giving rise to novel phenotypes for genomic selection and breeding purposes (e.g., change in biomass over time, growth, or senescence rate).

Uniformity is important because growers desire high forage biomass with even growth throughout a paddock and it is also a characteristic for determining plant breeder’s rights.

If plants in a forage cultivar are overly competitive, overall biomass yield and uniformity is expected be suboptimal in the paddock.