Deep Learning Models for the Classification of Crops in Aerial Imagery: A Review

Round 1

Reviewer 1 Report

Dear Authors:

      Your literature review article: “Deep Learning Models for the Classification of Crops in Aerial Imagery: A Review” presents a recurring theme that, despite the evolution of sensors and computational techniques, still presents challenges to be overcome. Regarding the writing of the article, the reading is fluid and the themes are well structured. However, I missed in chapter 1 some bibliographical references reporting the importance of the Classification of Agricultural Areas using remotely located data. The wording of the final version of the article must be reviewed by a native English speaker. Aiming to improve the article, I pointed out some minor comments that can be consulted in the digital file. I request your attention to the following questions:

1) Page 1: “Aerial imagery” The use of this terminology is controversial. Many bibliographical references refer only to images obtained by aerial and not orbital platforms. For example, the ISPRS definition for The term aerial imagery refers to all imagery taken from an airborne craft which can include drones, balloons or airplanes. And another definition:” Aerial and satellite imagery has been successfully utilized in a variety of areas such as topographic mapping, urban planning, assessment, forestry, precision environmental agriculture, water resources, and disaster monitoring. Terrain-based imagery (also referred to as close-range or terrestrial imagery) has seen a similar increase in prominence, making major inroads into mobile mapping, industrial metrology, forensics, cultural heritage preservation, medical imaging, underwater measurement and the gaming and movie industries in particular since the advent of unmanned aircraft systems (UAS) as new and flexible platforms. Such images are not used exclusively in the photogrammetric community, but also in many neighboring disciplines, in particular in computer science and electrical engineering, and typically under different nomenclatures, such as computer vision.

2) Pg 2: It would be interesting here to also present the concept of the data cube.

3) In Chapter 3 of Materials and Methods, the 4 objectives of the work are presented. It would be more appropriate to quote them at the end of chapter 2.

4) In the case of the AgriSAR mentioned in Table 1, it was developed with the following purposes:” The main outcomes of the AGRISAR 2006 project were:1) Preliminary analysis shows a high potential of using the acquired data for further studies, 2) Multidisciplinary approaches show a high potential of new product development and 3) SAR: depending on the application either C-band or L-band frequency is preferable. It actually served as the basis for the development of the Sentinel 1 and 2 platforms. Please explore this issue in the text.

5) In Chapter 5 you cited bibliographical references without presenting the metrics obtained by the authors. It would be interesting to standardize how presented for bibliographic reference [18]

6) Reference [24] is not complete.

    I conclude by congratulating them for the work done and for the presented version of the article.

Respectfully,

 

 

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

This research article focused on systematic review of deep learning models for the classification of crops in aerial imagery. The article is well flown. To improve the quality of the paper, the authors need to consider the following items for the revision.

 

Major comments:

1. It is suggested that the author revises and check English writing. Besides author check the author guidelines for the manuscript formatting.

2. In abstract, it is advised to add some major findings from the study and some key drawbacks/scope remain for further study.

3. The introduction is well flown. It is suggested to review the literature to add some more specific information. Such as: “aerial imagery” based on agricultural context is not sufficient. It needs some more technical clarification, same as Deep learning for classification.

4. The motivation of the study should go before section 3.  In section 3, more elaboration of article selection and search would be appreciable. Why author select between 2020 to 2022? Is there any specific reason?

5. More detailed about data acquisition process, data types, conditions during data acquisition such as light and other environmental factors can be added in the description during the crop classification methods explanations. Also sensors like camera specifications, image quality and constructed image information can be added.

6. Based on the addition (comment 5), more specific discussion should be added in the discussion. Geometric and radiometric calibration of aerial images also can be a factor for deep learning techniques. This can be included in the discussion section.

7. Revise the conclusion explaining the methods, findings and limitation of the DL methods with future scopes. I would like to suggest to move the answer part to the discussion section. This way, it would be more appropriate.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

The study reviewed the classification methodologies for the aerial images in agricultural field. The manusciprt is well written and acceptable for this journal.

However, I don't know why the authors uploaded their figures & tables in supplementary data. I think it should be checked in the editorial side.

Author Response

Please see the attachment.

Author Response File: Author Response.docx