Smarter Robotic Sprayer System for Precision Agriculture

Round 1

Reviewer 1 Report

1. The tile of this manuscript “Smart and Precision Electrical Sprayer for Agricultural Robots” is not closely related to the objectives listed in Line 40-47 and Line 100-103. In the tile, “Smart and Precision Sprayer” is a key issue whereas in the objectives “Agriculture robots” is. Please revise that.
2. In the manuscript, contents directly describes “Smart and Precision Electrical Sprayer” in 4. Electric based sprayer and 5. Crop perception system. But in 6. Tests and Results, it only describes results of crop perception system. It is suggested to add experiments and results about electrical sprayer in manuscript. It is also suggested to add results in manuscript about the join test in field or in laboratory for the two systems (Electric based sprayer and Crop perception system).
3. In Table 3, the SVM result is coming from training dataset or test dataset? Please describe it in text.
4. A stereo camera (two lenses) was used in here (Line 143) to record video. But the images analyzed in crop perception system were just one lens images. Please explain in text.
5. The crop conditions were classified into 0%, 33%, 66% and 100%. It is suggested to add a typical figure for these four conditions and explain their major differences.
6. The decision on crop condition classification is done by labors (Line 147). It is suggested to add texts in manuscript to describe the important criteria for their classification and also to explain why no instrument or device were used in here for the classification.
7. Please list out the company name and version for the SVM package (Line 175) in manuscript. Similar suggestion also goes to AgIot module in Line 114.

Author Response

1.The tile of this manuscript “Smart and Precision Electrical Sprayer for Agricultural Robots” is not closely related to the objectives listed in Line 40-47 and Line 100-103. In the tile, “Smart and Precision Sprayer” is a key issue whereas in the objectives “Agriculture robots” is. Please revise that.

Dear reviewer indeed the title is inducing the reader to the wrong work description. Taking this into consideration, we have updated the title to align with the work objectives described in Line 40-47 and Line 100-103.

2. In the manuscript, contents directly describes “Smart and Precision Electrical Sprayer” in 4. Electric based sprayer and 5. Crop perception system. But in 6. Tests and Results, it only describes results of crop perception system. It is suggested to add experiments and results about electrical sprayer in manuscript. It is also suggested to add results in manuscript about the join test in field or in laboratory for the two systems (Electric based sprayer and Crop perception system).

It were added in the tests and results section experiments that were performed with the sprayer, where the air flow was maintained and different references for water rate and centrifugal disk velocity were given and the results were observed.

3. In Table 3, the SVM result is coming from training dataset or test dataset? Please describe it in text.

The sentence was changed to “This procedure was repeated in all fourteen tests and the results obtained187were recorded on the test dataset (Table 3).” (Line 187)

4. A stereo camera (two lenses) was used in here (Line 143) to record video. But the images analyzed in crop perception system were just one lens images. Please explain in text.

The sentence was changed to “After, it was extracted images \added{of one of the lenses} from the recorded file, sampled every five frames to reduce the correlation between images and avoid annotating similar images.” (Line 145)

5. The crop conditions were classified into 0%, 33%, 66% and 100%. It is suggested to add a typical figure for these four conditions and explain their major differences.

It was added a figure with the four classes with the respective explanation: “ When there is no leaf in the region, it is considered 0%. When there are a few leaves, but in small quantities, never exceeding half of the total area, it is 33%. On the other hand, if a large part of the area is covered with leaves, but not totally covered, it is considered 66% and when the whole area is filled with leaves, it is considered a 100% class (Figure 6).”

6. The decision on crop condition classification is done by labors (Line 147). It is suggested to add texts in manuscript to describe the important criteria for their classification and also to explain why no instrument or device were used in here for the classification.

We have added an introduction in section 5.1 to better explain this issue.  

7. Please list out the company name and version for the SVM package (Line 175) in manuscript. Similar suggestion also goes to AgIot module in Line 114.

Some information was added about the SVM package and the AgIoT module.

Reviewer 2 Report

The manuscript presents the prototype of a mobile robot for precision spraying in vineyards. The technical description is easy to read and scientifically sound (e.g., the evaluation of the classifier).

However, it is not sufficiently clear what is the novel contribution of this work compared to the state of the art. There already is a lot of work on robotic precision spraying based on plant detection.

No evidence is given that the proposed spraying system enables the robot to treat the plants with the desired precision. Unlike other systems (which use robotic manipulators, for instance), the position and direction of the spraying devices cannot be adjusted in this system. Even the height of the spraying fans appears to be fixed, restricting the application of the robot to plants of corresponding height.

Furthermore, it remains unclear whether the robot is able to drive autonomously and what extent of operator intervention is required. The authors claim that the robot is able to drive on steep terrain, however, no evidence or quantification of this capability is given.

Regarding the SVM classifier, the rationale which has led to the definition of the four classes should be discussed, since alternatives are conceivable up to an estimator providing a continuous numerical value of the leaf density. Additionally, an evaluation of the classifier results on images showing other vegetation than the considered crop (e.g., on grassland) would be of interest.

There is a dangling link in the reference to the dataset [17]. Apparently, the publicly available data only consists of the recorded images. Including the annotations would make the dataset more useful for other researchers.

Author Response

1. However, it is not sufficiently clear what is the novel contribution of this work compared to the state of the art. There already is a lot of work on robotic precision spraying based on plant detection.

We have added a new table under state of the art to clarify this point. Indeed there are a huge number of works related to precision spraying, but mostly related to arable farms and not for woody crops. Woody crops brings some challenges to robotic sprayers.

 

2. No evidence is given that the proposed spraying system enables the robot to treat the plants with the desired precision. Unlike other systems (which use robotic manipulators, for instance), the position and direction of the spraying devices cannot be adjusted in this system. Even the height of the spraying fans appears to be fixed, restricting the application of the robot to plants of corresponding height.

In contrast to robotic manipulators, our approach allows more higher velocities during the spraying procedure. Besides our system has 3 independent sprayer set ( fan, atomizer, nozzle (pump) ) which are independently controlled and adjusted in height. Since they are controlled independently the system can reach good precision and reduce losses.

 

3. Furthermore, it remains unclear whether the robot is able to drive autonomously and what extent of operator intervention is required. The authors claim that the robot is able to drive on steep terrain, however, no evidence or quantification of this capability is given.

We added a set of references where is explained what robotic navigation is considered in this robot.

4. Regarding the SVM classifier, the rationale which has led to the definition of the four classes should be discussed, since alternatives are conceivable up to an estimator providing a continuous numerical value of the leaf density. Additionally, an evaluation of the classifier results on images showing other vegetation than the considered crop (e.g., on grassland) would be of interest.

We have selected 3 ROIs to feed independent controllers of the three sprayers set (fan, atomizer, nozzle ). These controllers needs to be feeded with leaf area index,  so we selected the use of a classifier with 4 classes, to reduce the system complexity. However, indeed you are right,  other alternatives are conceivable to obtain a more continuous values. But, from our perspective, a more continuous value will not bring to much benefits in terms of precision. The variability on the leaf area is not so high that justifies a more complex system to obtain a continuous variable. Typically, the majority of vines are fully healthy (with all leaves) or simply dead, the mid term is residual. Nevertheless, to consider this mid term we have added 2 more classes reach a more precision system. Some of this information was inserted in the Dataset section.

We have added in future work to consider and evaluate the need of a more continuous leaf area classifier (we can consider a linear regression formulation for the leaf area index).

Regarding testing the system with other vegetation crops is very relevant point, but we believe that the robot will be feeded with prescription map (with product application amount and  ) for operation on woody crops. The existence of other tree/plan should be not a problem because, if exist another plan specific location the prescription value should be 0. Nevertheless, the video presented on the article you can see a robot passing in a region without a vine and another tree in the background (please take a look at the second 6-10).

 

5. There is a dangling link in the reference to the dataset [17]. Apparently, the publicly available data only consists of the recorded images. Including the annotations would make the dataset more useful for other researchers.

The dataset contains the “images.zip” file, with the recorded images and the “annotatios.txt” file with the annotations.

Reviewer 3 Report

The authors developed a useful electric sprayer and included it in a mobile robotic platform. They calculated the leaf density based on a Support Vector Machine (SVM) classifier using image histograms (Local Binary Pattern (LBP), Vegetation Index, Averages and Hue). This density can then be used as a reference value to feed the controller that determines the airflow, the water rate and the water density of the sprayer. Experimental results demonstrated that the strategy worked well. This paper could be very interest of the researchers or engineers in this field. Before publication, please consider the following suggestion.

1. The contents were cut into too many paragraphs. Some figures should be combined.
2. The authors are suggested to provide a video to show how the robot works and the details of the electric sprayer under different conditions.
3. More recent research works should be added in the reference.
4. The section titles should be improved.
5. A table list the existing typical electric sprayers with their advantages and disadvantages could help the readership understand the new point of the design presented in this manuscript.
6. Section 3 and 4 should add more details, especially the design of the sprayer.

Author Response

1. The contents were cut into too many paragraphs. Some figures should be combined.

Figure 2 and 3 were combined.

2. The authors are suggested to provide a video to show how the robot works and the details of the electric sprayer under different conditions.

It was added in the results section a link to a video of a demonstration.

3. More recent research works should be added in the reference.

One reference was added: “Recently, a research group developed a solution based on a robotic vehicle for distributing plant protection products in vineyards and greenhouses. The system was equipped with a standard spraying machine that was transformed to optimise spraying operations.”

4. The section titles should be improved.

The titles of section 3 and 4 were changed.

5. A table list the existing typical electric sprayers with their advantages and disadvantages could help the readership understand the new point of the design presented in this manuscript.

There are very few works related to electrical sprayers based designs. This is an ongoing R&D line (in very recent years due to emerging electrical tractors/car technologies) mostly being realized by commercial companies. Since, there are few works that can be presented/referred on the state of the art, we choose to describe the general design for agricultural sprayers and present their advantages and disadvantages. We hope this brings more relevance to the article. (Table “Sprayer design approaches comparation”)

6. Section 3 and 4 should add more details, especially the design of the sprayer.

It were added some details, namely an electrical schematic of the entire system.

 

Round 2

Reviewer 1 Report

1. Line 235-248, centrifugal disk reference and water pump reference, what is the meanings of the reference? , speed, voltage, or others?, please explain in text.

 

2. In the title of Figure 9, please add texts to indicate corresponding water sensitive paper for test1 to test5 (Line 235-248).

 

3. The video link in Line 249, shows a video length 53’ 03”. Please extract the video within two minutes that directly relates to sprayer experiments.

 

4. The later part of the previous reviewer’s comments No.2 has not been answered properly. It is listed again as follows: “2. In the manuscript…. It is also suggested to add results in manuscript about the join test in field or in laboratory for the two systems (Electric based sprayer and Crop perception system).”

 

Author Response

1. Line 235-248, centrifugal disk reference and water pump reference, what is the meanings of the reference? , speed, voltage, or others?, please explain in text.

It was added a table with the reference values used (Table 6).

2. In the title of Figure 9, please add texts to indicate corresponding water sensitive paper for test1 to test5 (Line 235-248).

 

The following text has been added to the figure legend: “The papers represent the results of tests 1 to 5, from left to right, respectively.”

 

3. The video link in Line 249, shows a video length 53’ 03”. Please extract the video within two minutes that directly relates to sprayer experiments.

 

In fact the last video presented is too long and much of the content is not directly related to the sprayer experiments. The video has been replaced by a shorter version.

4. The later part of the previous reviewer’s comments No.2 has not been answered properly. It is listed again as follows: “2. In the manuscript…. It is also suggested to add results in manuscript about the join test in field or in laboratory for the two systems (Electric based sprayer and Crop perception system).”

 

As already mentioned, the two systems have not yet been tested simultaneously. Until now, the sprayer was built and its operation validated through the water sensitive paper, and the crop perception system was developed and validated with a dataset of vineyard images collected previously. This way we think it is better to put the test of the two systems as future work.

Reviewer 2 Report

The manuscript has improved substantially. The remaining concern is mainly the insufficient comparison to the state of the art.

Neither the text nor the table points out whether the "fully electric sprayer" approach has already been pursued by other researchers. It might be helpful to add the referenced systems to the categories of Table 1. Some aspects explained in the reply are not yet mentioned in the manuscript (e.g., woody crops; velocity). Altogether it is still difficult to identify the unique benefits of the presented system.

The DOI link to the dataset in reference [23] is still not working.

How does the "Controller system" in Figure 6 work? For instance, which reference values are adjusted according to the SVM classification results?

Layout and spelling should be checked again especially in the new sections. Table 1 has a very small font size and inconsistent capitalization.

I am not sure if Figure 3 is useful for the readers and if the notation is understandable. Perhaps a more abstract (high-level) illustration could be more helpful.

In the water-sensitive paper tests of Section 6, the experiments would perhaps be easier to understand if the reference values where given in tabular form, either quantitative or qualitative (low, medium, high, etc).

The video linked in line 249 is very long and it is not obvious where the topics of this article are covered. In my opinion, a much more concise video in the style of the video from line 225 would be more informative.

Author Response

The manuscript has improved substantially. The remaining concern is mainly the insufficient comparison to the state of the art.

 

1. Neither the text nor the table points out whether the "fully electric sprayer" approach has already been pursued by other researchers. It might be helpful to add the referenced systems to the categories of Table 1. Some aspects explained in the reply are not yet mentioned in the manuscript (e.g., woody crops; velocity). Altogether it is still difficult to identify the unique benefits of the presented system.

 

The majority of these solutions are already commercial solutions and there are a lot of suppliers. It will be harder to select a specific literature reference or a brand regarding this topic. Besides, the science literature focus on variable rate technologies and not on the sprayer hardware itself. Our paper proposes a different perspective on the innovation of sprayers concept, by redesigned the full system from scratch where, from our knowledge, we can not found more references than those presented already in the document.

 

2. The DOI link to the dataset in reference [23] is still not working.

 

The DOI was corrected.

 

3. How does the "Controller system" in Figure 6 work? For instance, which reference values are adjusted according to the SVM classification results?

 

The controller system is the junction of two systems, the electric based sprayer and the crop perception system. The two systems have not yet been tested simultaneously. Until now, the sprayer was built and its operation validated through the water sensitive paper, and the crop perception system was developed and validated with a dataset of vineyard images collected previously. This way we think it is better to put the test of the two systems as future work.

 

4. Layout and spelling should be checked again especially in the new sections. Table 1 has a very small font size and inconsistent capitalization.

 

Table 1 was corrected.

 

5. I am not sure if Figure 3 is useful for the readers and if the notation is understandable. Perhaps a more abstract (high-level) illustration could be more helpful.

 

In fact some details were presented in an inappropriate notation. Changes were made in figure 3, namely, in the AgIoT pins in order to have a more generic representation of the circuit.

 

6. In the water-sensitive paper tests of Section 6, the experiments would perhaps be easier to understand if the reference values where given in tabular form, either quantitative or qualitative (low, medium, high, etc).

 

It was added a table with the reference values used (Table 6).

 

7. The video linked in line 249 is very long and it is not obvious where the topics of this article are covered. In my opinion, a much more concise video in the style of the video from line 225 would be more informative.

 

In fact the last video presented is too long and much of the content is not directly related to the sprayer experiments. The video has been replaced by a shorter version.