Review Reports

Round 1

Reviewer 1 Report

First of all, I would like to congratulate the authors, since the mere fact of conducting a research is worthy of congratulations.
Even so, I have detected a high number of authors, which lowers the quality index of this article.

Next I will make a series of suggestions that could improve this publication.

Perhaps in the introduction it would be good to include some more bibliometric references.

In Figure 4, indicate the meaning of image a and b.

I see that it makes an exhaustive analysis of the techniques used in the identification of pests and diseases but I do not see the final purpose of this study.
If you try to identify which methods are most used perhaps if you introduce a Wordclouds with the purpose of the most determinant uses. This may be the final result of the study.

The conclusion of the study is not clear to me, you could redo this section.

First of all, I would like to congratulate the authors, since the mere fact of conducting a research is worthy of congratulations.
Even so, I have detected a high number of authors, which lowers the quality index of this article.

Next I will make a series of suggestions that could improve this publication.

Perhaps in the introduction it would be good to include some more bibliometric references.

In Figure 4, indicate the meaning of image a and b.

I see that it makes an exhaustive analysis of the techniques used in the identification of pests and diseases but I do not see the final purpose of this study.
If you try to identify which methods are most used perhaps if you introduce a Wordclouds with the purpose of the most determinant uses. This may be the final result of the study.

The conclusion of the study is not clear to me, you could redo this section.

Author Response

Firstly, I sincerely appreciate your thorough feedback on this article. Secondly, your point about the potential impact of the large number of authors on the quality of the article is indeed a valid consideration. The paper demands extensive reading and analysis of literature, and each mentioned author has contributed in varying degrees to this article, whether writing, providing constructive feedback on the structure and content, supplementing with additional references, and so on. If the number of authors surpasses the limit, discussions could be held with the latter authors to determine potential deletions. Finally, I have made revisions and provided explanatory notes based on your valuable suggestions.

1. Regarding the issue of limited references in the introduction, a careful analysis of the introduction has been conducted, and new references have been incorporated to provide a more substantiated content.
2. Lines 255-272 of the paper are dedicated to the analysis of Figure 4. This analysis is divided into two main sections. The first section examines the types of crops mentioned in the relevant literature, while the second section focuses on the types of pests and diseases associated with several key crops. Following the revisions, both Figure 4a and Figure 4b are explicitly referenced in both the figure caption and the main text.
3. Finally, an explanation is provided for the paper's objectives, identification methods, and conclusions. In the first section of this paper, the writing purpose is clarified (lines 51-53: intelligent identification of regions with crop pests and diseases for targeted spraying, reducing pesticide usage, and enhancing pesticide efficiency). The main contents of the paper are summarized around this objective. The second section briefly outlines the potential applications of different airborne remote sensing imaging sensors in agricultural pest and disease identification. Building upon a wealth of relevant literature, the third section presents a comprehensive analysis of techniques for identifying crop pests and diseases, encompassing image acquisition processes, feature extraction methods, discriminative algorithms, and more. The fourth section summarizes commonly used methods for identifying crop pests and diseases, which can serve as references for research on low-altitude drone-based pest and disease identification (as depicted in Figure 6). The section also highlights potential issues with these methods in practical applications (such as the observation that most data involving pest and disease images are sourced from artificially infected plots, indicating controlled stress conditions in crops, which might not fully represent real-world scenarios). Therefore, a more universally applicable approach for identifying crop pests and diseases is proposed, illustrated in Figure 7. To align with the objective outlined in the first section, extensive reading and analysis have led to the proposal of identification methods more suitable for naturally occurring pest and disease conditions. This stands as the significant conclusion of this paper.

Author Response File: Author Response.docx

Reviewer 2 Report

The authors comprehensively articulated the applications of UAVs in the identification of crop pests and diseases. A few suggestions are as below:

- Avoid abbreviations in the abstract if you have to may they be expanded.

Line 244: Can change "home" to "in China" to make the context suitable for the international audience rather than just for China

Line 289: Delete "will"

Line 308: Change "focuses" to "focused"

It is ok in my view

Author Response

Thank you for your thoughtful feedback on the article. 

Regarding the first suggestion to " avoid abbreviations in the abstract," I have made the necessary revisions in the paper. In the abstract, "unmanned aerial vehicles (UAVs)" holds particular importance, with an occurrence of more than 3 times. Therefore, I have opted to refrain from using the abbreviation for the initial instance and have subsequently replaced it with the abbreviation. The same approach has been applied to "UAV Low-Altitude Remote Sensing (UAV-LARS)."

Secondly, in response to the grammatical errors identified in Line 244, Line 289, and Line 308, I sincerely appreciate your corrections, and I have already made the necessary revisions.

Author Response File: Author Response.docx

Reviewer 3 Report

The paper presents a review on applications of UAV-LARS systems in agricultural processes (in both research and industrial directions). I cannot say that this review is very soundful and may really influence the community but the research is not misleading or useless. However, my main sphere of professional interest is related to the computing and modeling technologies which are almost not discussed in this manuscript.

 

In this paper authors discuss a lot of detail about the available sensors (including the wavelenth of their channels) and their technical issues. However, authors do not pay much attention to the algorithmic back-end (even though the methods are listed in the tables) of their applications and it is understandable -- it is hard to pack all issues into a single review. Anyways, I would like to encourage authors to provide more extensive discussion about the algorithms in the revised text.

I may add that sometimes the satellite remote sensing technologies may be useful for detection of low-production field areas (see e.g. paper Journal of Physics: Conference Series 1117, 012009) even though authors discuss that low-altitude remote sensing leads to better understanding of the target processes.

Is it necessary for this reserach area to exploit only the flight platform but not the array of static sensors for solving the chosen target task? See e.g.

-- Computer vision-based platform for apple leaves segmentation in field conditions to support digital phenotyping. Computers and Electronics in Agriculture, 201, 107269. (2022) -- Dynamic Mode Decomposition and Deep Learning for Postharvest Decay Prediction in Apples. IEEE Transactions on Instrumentation and Measurement. (2023)   and related papers.  

I think that technical documentation about the listed software (in Section 5.2) should be added to the reference list and cited in a proper way.All in all, I see that the paper requires a revision and additional review by the person with more knowledge on technical side of the discussed technology review.

Author Response

Thank you very much for the professional advice you mentioned. I have made revisions and provided explanatory notes based on your valuable suggestions.

1. This manuscript extensively analyzes the techniques for crop pest and disease identification based on a large number of relevant literature. In particular, a significant amount of research has been conducted on image acquisition and feature extraction. Common algorithms such as Support Vector Machines, Random Forests, and Linear Regression are frequently used for model construction, so there has been no in-depth discussion.
2. The effective acquisition of the target dataset is a prerequisite for building high-precision recognition models. Section 4 of the manuscript mentions that most of the literature uses target datasets from artificially inoculated plots, indicating that the stress conditions of the crops are controllable, greatly reducing the difficulty of dataset acquisition. The extraction of sensitive features is beneficial for accurately distinguishing different pests and diseases as well as quantifying stress levels. Different crops and different pests and diseases have different sensitive features, including vegetation indices, spectral indices, spectral bands, and texture features, among dozens of other feature parameters. It is necessary to summarize these features to quickly select appropriate feature parameters in the research. Regarding the backend algorithms, discriminative algorithms that can achieve higher accuracy mainly include logistic regression, linear regression, support vector machines, and other commonly used methods, which are listed in Section 4. From Tables 1 and 2, it can be observed that these algorithms are not specifically designed for the recognition of a particular pest or disease (even considered as general, such as SVM), therefore, no further in-depth discussion has been conducted on them.

3. Firstly, I fully agree that ground-based remote sensing and satellite remote sensing technologies are beneficial for yield assessment and even for monitoring and identifying crop pests and diseases. Secondly, regarding the two referenced articles mentioned, the identification of infected apple trees and the monitoring of damages and diseases occurring during apple storage are achieved using static sensors to obtain high-resolution target task datasets, which can lead to higher recognition accuracy. Additionally, the algorithm models can be transferred to mobile devices, facilitating the inspection of disease conditions for each individual tree. This method is also effective for identifying infected crop leaves, but it is less efficient for guiding the overall pesticide application in the entire field. Finally, compared to static sensors, the advantage of drones lies in their ability to monitor agricultural fields on a large scale, acquiring images of the entire field within a relatively short period of time. However, the high efficiency of obtaining agricultural information sacrifices ground spatial resolution and increases the difficulty of subsequent processing. This is why it is necessary to conduct specialized research on the low-altitude remote sensing of crop pests and diseases using drones.

4. Regarding the software mentioned in section 5.2, references have been added.

Author Response File: Author Response.docx