Rice Height Monitoring between Different Estimation Models Using UAV Photogrammetry and Multispectral Technology

Round 1

Reviewer 1 Report

Drones are a most useful new tool in crop estimations and this paper provides an important analysis of drone use.  However, the analysis and consequent results are complex, and so there needs to be some improvement in the presentation of the methods used and results. 

Comments for author File: Comments.pdf

Author Response

Dear Editor and Reviewers

 

Thank you for your letter and for the reviewers’ comments concerning our manuscript entitled “Rice height Monitoring between different estimation models using UAV photogrammetry and multispectral technology” (remotesensing-1519909). We have changed the title to: “Rice height monitoring between different estimation models using UAV photogrammetry and multispectral technology”. Those comments are all valuable and very helpful for revising and improving our paper. We have studied comments carefully and have made correction which we hope to meet with approval. The revised portion are marked in red. The main corrections in the paper and the responses to the reviewers’ comments are as following:

 

Reviewer: 1

【Comments】

Drones are a most useful new tool in crop estimations and this paper provides an important analysis of drone use.　However, the analysis and consequent results are complex, and so there needs to be some improvement in the presentation of the methods used and results.

 

【Revision】

The methods and results of this study correspond to each other. We have restructured and rewritten the Results and discussion sections.

 

【Comments】

One general comment: many abbreviations used here are relatively standard, but others are less so.  This paper will be read not only by experts in the field who know what the abbreviations mean, but by others interested in new applications of drone technology.   As a result, I suggest first of all that all abbreviations be spelt out in all titles of sections:

eg line 289: 2. 5. Extraction of point clouds and establishment of Crop Height Models (CHM) and measured height (MH) 

And I also suggest that if an abbreviation has not been used for more than a page, it should be spelled out: eg line 256: spell out DSM and DTM in line 256, SPAD in line 435

 

【Revision】

We rechecked all the abbreviations and changed them accordingly.

 

【Comments】

Specific comments and suggested alterations:

Abstract:

Line 10-12: Abstract: Unmanned aerial vehicle (UAV) photogrammetry was used to monitor crop height in a flooded paddy field. Three multi-rotor UAVs NOT… UAVs (UAVs) were utilized to conduct flight missions to capture RGB (RedGreenBlue) and multispectral images and these images were analyzed using several different models to provide the best results.

 

【Revision】

We have modified.

 

Introduction:

【Comments】

Line 34: ..so an adequate rice…

 

【Revision】

 We have revised.

 

【Comments】

Line 38: genetic factors [2], which is a critical….

 

【Revision】

We have revised.

 

【Comments】

Line 54: …time-consuming and labor-intensive work when areas of crop are large [10], but…

 

【Revision】

We have revised.

 

【Comments】

Line 60: ..in precision agriculture…

 

【Revision】

We have revised.

 

【Comments】

Line 63-64: …provide a wider area of assessment at the same time…

 

【Revision】

 We have revised.

 

【Comments】

Line 65: …but also non-invasive and non-disruptive.

 

【Revision】

 We have revised.

 

【Comments】

Line 73: ..human activities within the crop must of necessity cause some disruption and damage.

 

【Revision】

We have revised.

 

【Comments】

Line 88: remove “at present” 

 

【Revision】

We have deleted.

 

【Comments】

Lines 112-115: have a space between words and references eg layer [36] and in line 115.

 

【Revision】

We have deleted the space.

 

【Comments】

Materials and Methods:

Line 144: The experimental field is a rectangular area…

 

【Revision】

We have revised.

 

【Comments】

Line 146: I do not think 15 decimal places are needed for the latitude and longitude.

 

【Revision】

We have replaced with WGS-84 coordinates as follows:

Line 143-144: …Ami town, Ibaraki prefecture of Japan (36°01'58.0"N 140°12'42.9"E)…

 

【Comments】

Line 153: A plot consists of 2 areas.

 

【Revision】

We have revised.

 

【Comments】

Figure 1b. Please make the sampling area squares much darker. Also: explain what is the meaning of the numbers 15608301>>15608460 and 4304320>>4304640. Why not use latitude and longitude?

 

【Revision】

“15608301>>15608460 and 4304320>>4304640” is the coordinate format based on projected coordinate system (WGS 84 / Pseudo-Mercator)

We revised and reformat as WGS 84 (EPSG: 4326) coordinates

 

【Comments】

Table 1: Change “Availability of cover crop” to “Cover crop present” and then No or Yes 

(A1 No A2 Yes A3 Yes A4 No etc…)

 

【Revision】

We have changed.

 

【Comments】

Line 166: 114cm between rows for dense sowing, 26 cm between rows for sparse sowing

 

【Revision】

We have revised.

 

【Comments】

Line 169: remove “respectively” at end of line

 

【Revision】

We have removed it.

 

【Comments】

I suggest using italics to emphasize the differences in the drones:

Line 170-174: P4P is a consumer-grade drone equipped with a DJI FC6310 camera…… Mounted on the P4P was a LiPo 2S battery with 6000mAh capacity to support flight for about 30 minutes (Table 2).

Line 174: P4R is also a consumer-grade drone that has an increased hover accuracy (Table 2) and is equipped with a DJI FC6310R camera.

Line 176: P4M is a dedicated and customized UAV drone for plant or crop monitoring through a series of built-in sensors.

 

【Revision】

We have revised.

 

【Comments】

In both Table 2 and 4 the current alignment of Items made reading them very confusing.

For Table 2, I suggest aligning all items to the side thus:

Weight (g) including battery & propellers

Max flight Time (minutes)        ~30   ~30  `~27   (~ = approx.)

Hover Accuracy….

Hover Accuracy…

Battery

Operation ground station  

Camera:     Sensor

        Effective Pixels

        Focal Length

        FOV of lens 

        Image Size

        Photography format

 

Similarly in Table 4:

Image Quality Estimation

Accuracy

Pre-Alignment

Alignment:    Reference preselection mode 

          Key point limit

          Tie point limit

Camera Calibration

Etc

 

【Revision】

We have revised and reorganized the tables. In addition, the Table 2 and 4 were moved to supplementary material as Table S1 and Table S3, respectively.

 

【Comments】

Line 204: …all camera lenses for the three UAVs were corrected for distortion measurement during the production process.

 

【Revision】

We have revised.

 

【Comments】

Line 212: flight missions were conducted six times……

 

【Revision】

We have revised.

 

【Comments】

Line 214: ..SPAD values, were measured manually…    (Data is plural)

 

【Revision】

We have revised.

 

【Comments】

Line 222: The whole operation was managed by….

 

【Revision】

We have revised.

 

【Comments】

Figure 3, Line 247: Put *footnote on next line thus:

….and export of the dense point cloud was performed in sequence.

*Calibration method of P4R-based image processing in Metashape which is a …. 

Also: should the * be after P4R* and not P4P* as it currently is in the Figure?

Also in Figure 3: Align photos, Gradual selection Build dense cloud do show up clearly but why not make  them  stand  out  more  by  using  capitals: ALIGN  PHOTOS     GRADUAL  SELECTION   BUILD DENSE CLOUD

 

【Revision】

We have revised and then reorganized the Figure 3.

 

【Comments】

Line 256: spell out so to remind the reader what they are:   … such as Digital Surface Model (DSM) and Digital Terrain Model (DTM), are typically assessed by the deployment of Ground Control Points (GCPs).

 

【Revision】

We have revised.

 

【Comments】

Line 289: spell out what CHM and MH are in the title of the section

 

【Revision】

We have revised as follows:

2. 5. Extraction of point clouds and establishment of Crop Height Model (CHM) for Measured Height (MH) estimation

3.2. The results of the performance of the two Crop Height Models (CHMs)

3.4. The performance of the M3 method for Measured Height (MH) estimation

 

【Comments】

Line 324:   …as digital cloud points for the waterway and the rice canopy in DSPC,….  

 

【Revision】

We have revised.

 

【Comments】

Line 404: ..grayscale pixels of a Region of Interest (ROI) representing….

 

【Revision】

We have revised.

 

【Comments】

Line 435: Again, because it has been defined many sections ago, say what SPAD is:

The Soil Plant Analysis Development (SPAD) of the rice leaf blade is another set of data…..

 

【Revision】

We have revised as follows:

2.6.2. The potential of NDVI_canopy  and Vegetation Fraction (VF) and Soil plant Analysis Development (SPAD) value for Measured Height MH estimation

Line 425-426: …The Soil Plant Analysis Development (SPAD) of the rice leaf blade is another set of data obtained…

 

In addition, some of the VF and VI related title sentences have been changed as follows:

2.5.2. Vegetation Index (VI) filter and the equations for the M1 and M2 methods

2.6.1. Vegetation Fraction (VF) and canopy Vegetation Index (VI) creation

 

【Comments】

Figure 12 might be clearer if Measured Height (m) was placed beside 12a and beside 12c

 

【Revision】

We have reformated the Figure 12 (now is Figure 9).

 

【Comments】

Line 458: From the results shown in Figure 12, ….

 

【Revision】

We have revised.

 

【Comments】

Line 461: .. is also positive at 0.80,     is now the same as in Figure 12b

 

【Revision】

We have revised.

 

【Comments】

Line 481-482: …independent and dependent variables…

 

【Revision】

We have revised. 

 

【Comments】

Line 488:   ..(3) 24-fold cross-validation…

 

【Revision】

We have revised.

 

【Comments】

Results and Discussion

Line 502: …and theoretical coordinates

 

【Revision】

We have revised.

 

【Comments】

Line 513-515: ..(a)-(c) are the bias in XY, Z and XYZ levels based on Phantom 4 Pro V2.0. (d)-(f) are the bias in XY, Z and XYZ levels based on Phantom 4 RTK.

 

【Revision】

We have revised.

 

【Comments】

Line 523: …between the two UAVs (P4P and P4R) in images from different dates set in …..

 

【Revision】

We have revised.

 

【Comments】

In Table 5, why not use cm for P4P, then “The high error of P4P-based images” is obvious

 

【Revision】

We have revised and changed the unit.

 

【Comments】

Line 534: 3. 2. The results of the performance of the two Crop Height Models (CHMs)

 

【Revision】

We have revised.

 

【Comments】

Line 541+: ..The specific manifestations of different treatments: (a) low density without green fertilizer; (b) low density with green fertilizer etc

 

【Revision】

We have revised.

 

【Comments】

Line 552: ..based on P4P had high linear correlations where R2 was 0.9223 … .. but it still was 0.8844. Comment on lines 552-594: all of the correlations are relatively high, with some very high showing the value of the techniques tested here.

 

【Revision】

We have revised.

 

【Comments】

Line 598: ..correlation between estimated Crop Height (CH) and Measured Height (MH) on plots with different treatments…

 

【Revision】

We have revised.

 

【Comments】

Then in Figure 16, Change “Estimated Height (m)” to “Crop Height (m) Estimated From CHM” so that it is clear that you are comparing CH and MH mentioned in lines 598-602.

 

【Revision】

We have revised.

 

【Comments】

Figure 17 (a) The Aug.25.2020 is hard to read: the lettering needs to be darker

 

【Revision】

We have reformated the Figure 17 (a) (now is Figure 15(a))

 

【Comments】

Figure 17 (b) do not use “shows the ASS” as that is rather rude in English   Therefore, write “absolute spatial stability” in full in line 625, Figure 17 (b), line 644, line 655

 

【Revision】

We have revised.

 

【Comments】

Line 658: .. were used as inputs for the MLR model…

 

【Revision】

We have revised.

 

【Comments】

Line 671: ..MH data in the separate linear regression….

 

【Revision】

We have revised.

 

【Comments】

Line 675: …more validations of this model are needed.

 

【Revision】

We have revised.

 

【Comments】

Line 697: ..MH in each separate plot and in the whole area.

 

【Revision】

We have revised.

 

【Comments】

Lines 702-705:   ..different treatments, high R2 of between 0.8484 to 0.8628 were obtained for most treatments, with high density with green fertilizer having a lower R2 of 0.7887 (Figure 18).

 

【Revision】

We have revised.

 

【Comments】

Line 708: ..for MH estimation of the whole area.

 

【Revision】

We have revised.

 

【Comments】

Conclusions:

The conclusions are very important so spell out what CHMs MLR, DSPC, DTPC, SPAD, MH are

 

【Revision】

We have rewritten the conclusion section to make it more concise and straightforward as follows:

 

4. Conclusions

The ground where flooded rice grows is difficult to be directly detected by UAV-based photogrammetry. This study discussed the performance of consumer-grade UAV (P4P), professional-grade UAV equipped with a high-precision receiver (P4R), and dedicated UAV (P4M) for multispectral analysis to estimate CH of flooded rice during the growing season from different perspectives. All methods are feasible, which provides more possibilities in crop height estimation. Although the results with high are obtained through these technologies, there are still some improvements needed. For the M1 and M2 methods, the next target is to implement the integrated and automated processes from capturing images to creating CHM model, instead of stepped and cumbersome semi-automation. For the M3 method, to further discuss whether there was an inherent relationship between such different phenotypes, i.e., height and multispectral information, more calibration and validation for the present model are necessary. In addition, modeling crop height directly from more spectral information is a potential study.

 

【Comments】

Line 727: However, the results showed …

 

【Revision】

We have revised.

 

【Comments】

Acknowledgments:

Line 747: …for their help in relation to data collection…..

 

【Revision】

We have revised.

Author Response File: Author Response.pdf

Reviewer 2 Report

The ms remotesensing-1519909 entitled Rice height Monitoring between different estimation models using UAV photogrammetry and multispectral technology is well organized and written. In this study, the flooded paddy field was applied to monitor the crop height using different modeling through unmanned aerial vehicle (UAV) photogrammetry. Below are my comments, and the authors should revise them to make their ms stronger and suitable for publication in Remote Sensing.

Title: why the first letter in Monitoring is capital?

Keywords: remove unmanned aerial vehicle. Also, authors should start the keywords with capital letters.

L34-35: please cite the relevant and recent reference in the first sentence: https://doi.org/10.3390/plants10081657

https://doi.org/10.1007/BF00989136

L43: Do not start the sentence with abbreviations CCC even if you defined it before. The same issue for CH, please add the full words in the beginning of the sentence.

L95-97 who said this, please add relevant citation

L115-123 again, add relevant citation please

L123: Please do not start the sentence with such format of citations [36,39]. You should add the name of the author before these tow citations. For example, Malambo et al. [36] and xxx [39] used …

Table 1: please leave space between the number and the unit, see 26cm. Please, revise this issue within the ms.

L214: how did author measured leaf SPAD values? Which equipment they used? Add the model name and company and country of making.

The authors should add references for any methods they used within the material and methods section.

Results section was well resented, while the discussion part was poorly presented. The authors have to make discussion stronger than the current form. Also, I suggest authors to move some figures into supplementary material.

The conclusion, the authors should make it shorter and straight. They should only present the most important findings in this section.

 

Best regards, Reviewer

 

 

 

Author Response

Dear Editor and Reviewers

 

Thank you for your letter and for the reviewers’ comments concerning our manuscript entitled “Rice height Monitoring between different estimation models using UAV photogrammetry and multispectral technology” (remotesensing-1519909). We have changed the title to: “Rice height monitoring between different estimation models using UAV photogrammetry and multispectral technology”. Those comments are all valuable and very helpful for revising and improving our paper. We have studied comments carefully and have made correction which we hope to meet with approval. The revised portion are marked in red. The main corrections in the paper and the responses to the reviewers’ comments are as following:

 

Reviewer: 2

 

【Comments】

The ms remotesensing-1519909 entitled Rice height Monitoring between different estimation models using UAV photogrammetry and multispectral technology is well organized and written. In this study, the flooded paddy field was applied to monitor the crop height using different modeling through unmanned aerial vehicle (UAV) photogrammetry. Below are my comments, and the authors should revise them to make their ms stronger and suitable for publication in Remote Sensing.

Title: why the first letter in Monitoring is capital?

 

【Revision】

We have changed the title to: “Rice height monitoring between different estimation models using UAV photogrammetry and multispectral technology”

 

【Comments】

Keywords: remove unmanned aerial vehicle. Also, authors should start the keywords with capital letters.

 

【Revision】

We have removed the “unmanned aerial vehicle” and revised the rest of the keywords. The keywords are now shown below:

 

Keywords: Flooded paddy field; Photogrammetry; Crop height; Global navigation satellite system; Multispectral; RGB

 

【Comments】

L34-35: please cite the relevant and recent reference in the first sentence: https://doi.org/10.3390/plants10081657

 

https://doi.org/10.1007/BF00989136

 

【Revision】

We have added the references.

 

【Comments】

L43: Do not start the sentence with abbreviations CCC even if you defined it before. The same issue for CH, please add the full words in the beginning of the sentence.

 

【Revision】

We have revised it as follows:

 

Line 42-47: … Leaf chlorophyll content is one of the essential indicators to evaluate the growth status of crops and can be used to understand the crop environmental stress and the level of N content [6,7]. Crop canopy cover is another important FBP that can indicate the crop emergence or the senescence status of certain crops [8]. Crop height is an important indicator, and a fundamental phenotypic parameter of crops, and [9] has shown the correlation between CH and wheat biomass …

 

【Comments】

L95-97 who said this, please add relevant citation

 

【Revision】

We have added the reference as follows:

Bannari et al., 1995 (DOI:10.1080/02757259509532298)

 

【Comments】

L115-123 again, add relevant citation please

 

【Revision】

We have deleted the sentence.

 

【Comments】

L123: Please do not start the sentence with such format of citations [36,39]. You should add the name of the author before these tow citations. For example, Malambo et al. [36] and xxx [39] used …

 

【Revision】

We have revised the sentence as follows:

 

Line 120-121 Malambo et al. [36] and Jimenez-Berni et al. [40] used a non-cropped field.

 

【Comments】

Table 1: please leave space between the number and the unit, see 26cm. Please, revise this issue within the ms.

 

【Revision】

We have revised.

 

【Comments】

L214: how did author measured leaf SPAD values? Which equipment they used? Add the model name and company and country of making.

 

【Revision】

We have added information about the device as follows:

 

Line 152-153 … soil plant Analysis Development (SPAD) value (measured by SPAD-502Plus, Konica Minolta) …

 

【Comments】

The authors should add references for any methods they used within the material and methods section.

 

【Revision】

We have added the reference as follows.

In 2. 4. Section: Line 270… coordinates. Root Mean Square Error (RMSE) [44] …

In 2. 5.1. Section: Line 328… to segment the incomplete water surface cloud points in DSPC and then fit these points using a 2.5D quadric approach [47] to create…

In 2. 5.2. Section: Line 340-341… fields were flooded. So, the DSPCs have not only canopy digital points but also water surface digital points (Figure 4c). All points…

Line 346 …effects. VARI is defined as shown in equation 5 [49].

In 2.6.1. Section: Line 377-378 …to the object. The NDVI is calculated as shown in equation 8 [52].

Line 385 VF was defined as the percentage of vegetation within a certain area [53,53] on…

Line 398 …To quantify VF and NDVI_canopy, regions of interest (ROI) (0.3 m×28 m) [54,55]

 

In 2. 7.2. Section: Line476-477 …validation groups, and then k-fold cross-validation was used to validate the stability of the MLR model [38]

 

【Comments】

Results section was well resented, while the discussion part was poorly presented. The authors have to make discussion stronger than the current form.

 

【Revision】

We have added a section (3.5 Evaluation and discussion based on the M1 and the M2 and the M3 methods) to 3. Results and disccussion and moved the original discussions that existed in the other sections to 3.5. In addtion, we have rewritten (added and removed) the contents of the discussion.

 

The specific changes are as follows:

 

In 3.1. section:

Line 525-533 were deleted.

Some contents were added to the end of Line 522 as follows:

… and vertical and total spatial level of two UAVs were calculated manually, and the results are shown in Table 2. The high error of the P4P-based images was expected and reasonable due to the limitation of the DPGS technique. The image error of this study was very close to the absolute positioning error of DGPS (~50cm) and RTK-GNSS (~3cm) receiver mounted on P4P and P4R, respectively. In addition, similar with the results above, the image set obtained on Oct.25.2020, when there was a bare paddy field after harvesting, had a lower error than other image sets obtained on other dates when there was rice growth…

 

In 3.2.1. and 3.2.2. section:

Line 562-572 and Line 589-595 were revised and moved into 3.5 section (Line 628-641) as follows.

On the performance of the models with different treatments, the M1 method obtained the highest correlation in the treatment of low density with cover crop and the lowest correlation in the treatment of low density without cover crop on both P4P and P4R cases. The plot with low planting density and growth of green fertilizer crop in waterway had a wider distribution of vegetation (Figure 13. b) and had fewer homogeneity of vegetation per unit area compared to high planting density (Figure 13. d). In plots with cover crop, low density treatments showed better , which may be influenced by homogeneity [64], while in plots without cover crop, high density treatment showed higher , which may be affected by the percentage of water region, and the previous studies have also indicated that the water surface can have an impact on the accuracy of photogrammetry [65]. Like the M1 method, in the M2 method, the treatment of low density without cover crop with a wider water region may lead the higher uncertainty during UAV photogrammetry, and a higher density of vegetation in high density with cover crop may also result in lower accuracy due to high homogeneity of vegetation.

 

In 3.2.3. section:

Line 612-634 were revised and moved into 3.5 section (Line 642-683).

 

We also added some contents into discussion (Line 684-700):

Unlike the comparison between height information in the M1 and M2 methods, the M3 method uses spectral information to estimate crop CH indirectly. Compared to the single factors discussed in 2.6.2., the model after multiple linear regression had better performance in estimating the overall crop CH in the paddy field. SPAD value is calculated by red and infrared light transmission, which is dedicated for chlorophyll content [68]. NDVI is then calculated from read and near- infrared light reflectance, which is designed for vegetation detection. So, it can be assumed that the trait CH relates to leaf chlorophyII concentration and overall NDVI of canopy. Nevertheless, compared to the results obtained by the M1 and M2 method, the   may be relatively low, which also implies that there may be more than two factors influencing the CH from the spectral level. However, On the performance with different treatments, except for the treatment of high density with cover crop, which showed a poor correlation, the remaining three treatments showed higher  than the overall, close to the other two methods. It is assumed that more complex vegetation characteristics of the high density plot with covercrop let higher uncertainty to target areas through UAV images. In addition, at the spectral level, the M3 method has introduced red, infrared, and near-infrared lights into the estimation of CH, so developing a new spectral model dedicated to crop height estimation may be the next.

 

【Comments】

Also, I suggest authors to move some figures into supplementary material.

 

【Revision】

We have move Table 2, Table 3, Table 4, Figure 8, Figure 9, and Figure 13 into supplementary material.

 

【Comments】

The conclusion, the authors should make it shorter and straight. They should only present the most important findings in this section.

 

【Revision】

We have rewritten the conclusion section to make it shorter and straight as follows:

 

4. Conclusions

The ground where flooded rice grows is difficult to be directly detected by UAV-based photogrammetry. This study discussed the performance of consumer-grade UAV (P4P), professional-grade UAV equipped with a high-precision receiver (P4R), and dedicated UAV (P4M) for multispectral analysis to estimate CH of flooded rice during the growing season from different perspectives. All methods are feasible, which provides more possibilities in crop height estimation. Although the results with high are obtained through these technologies, there are still some improvements needed. For the M1 and M2 methods, the next target is to implement the integrated and automated processes from capturing images to creating CHM model, instead of stepped and cumbersome semi-automation. For the M3 method, to further discuss whether there was an inherent relationship between such different phenotypes, i.e., height and multispectral information, more calibration and validation for the present model are necessary. In addition, modeling crop height directly from more spectral information is a potential study.

 

 

Best regards,

authors

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The ms has been improved, but the authors have to check the english for the text.