High-Resolution Remote Sensing Image Change Detection Method Based on Improved Siamese U-Net

Round 1

Reviewer 1 Report (Previous Reviewer 1)

 General Comments

The first part of the manuscript is comprehensible and scientifically valid. The introduction is well done and provides a clear overview of the state of the art. The objective is well explained, although the bullet point list could be removed. The applied methodology is well explained, and the adopted experimental design is correct.

However, there are some things that need further adjustment to improve the paper. Results appear to be more linear, but there is a lack of basic statistics to enhance understanding and assess the actual differences obtained. Unfortunately, there are not significant comments in the discussion section. In fact, there is a complete absence of comparison with other studies, and there is not even a citation.

 Manuscript proposed a novel approach to address the challenges associated with high-resolution
remote sensing image change indistinct detection boundaries, poor detection of small targets and
increased occurrences of pseudo changes. Manuscript use a classical semantic segmentation method
implemented with a method based on the Siam-FAUnet network. Furthermore, the method included
an ASSP module with variable expansion rates and a FAM module to address the problem of semantic
mismatch. The work exploits some deep learning algorithms and evaluates the level of accuracy of
the predictive models. Overall, the work focuses on an emerging research topic and exploits some
very promising image analysis techniques.
The article addresses a highly important topic for current research, namely the classification of high-
resolution images. However, the article is poorly structured in its various sections. Introduction’s
manuscript is comprehensible and scientifically valid. Indeed, it is well done and provides a clear
overview of the state of the art. The objective is well explained, although the bullet point list could
be removed. The applied methodology is well explained, and the adopted experimental design is
correct. However, there are some problems that need further adjustment to improve the paper. These
problems are correlated with the statistic applied, results obtained and their discussions. Results
appear to be more linear, but there is a lack of basic statistics to enhance understanding and assess
the actual differences obtained. It seems that the applied statistics are solely based on comparing
different classification algorithms across various datasets. More in-depth statistical analyses are
needed to truly understand which algorithms perform better and under which conditions.
Additionally, all the images appear to be very small, so they cannot be properly observed; in particular
the figure 8. Unfortunately, there are not significant comments in the discussion section. In fact, there
is a complete absence of comparison with other studies, and there is not even a citation.
Therefore, the manuscript requires a major revision.

Therefore, the manuscript requires a major revision.

General Comments

 

The first part of the manuscript is comprehensible and scientifically valid. The introduction is well done and provides a clear overview of the state of the art. The objective is well explained, although the bullet point list could be removed. The applied methodology is well explained, and the adopted experimental design is correct.

However, there are some things that need further adjustment to improve the paper. Results appear to be more linear, but there is a lack of basic statistics to enhance understanding and assess the actual differences obtained. Unfortunately, there are not significant comments in the discussion section. In fact, there is a complete absence of comparison with other studies, and there is not even a citation.

 

Therefore, the manuscript requires a major revision.

Author Response

Thank you very much for your time involved in reviewing the manuscript and your very encouraging comments on the merits. We also appreciate your clear and detailed feedback and hope that the explanation has fully addressed all of your concerns. Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report (New Reviewer)

The content of the paper appears to be well-presented and coherent.  However, as a reviewer, I would like to provide some feedback and suggestions to enhance the quality of the paper further. My comments are as follows:

The abstract lacks a clear statement of the research problem. It would be helpful to explicitly state the specific challenges faced in high-resolution remote sensing image change detection that the proposed algorithm aims to address.

Introduction: Consider expanding the section on the limitations of existing research to provide more context. What are the drawbacks of current change detection methods? Are there specific areas or scenarios where the limitations are particularly pronounced? This information would help justify the need for a novel algorithm like Siam-FAUnet.

L219: develop: *typo

While the contributions of the research are mentioned in the introduction section, they could be presented more explicitly. Consider revising the section to clearly outline the key contributions of the proposed Siam-FAUnet algorithm, such as its novel combination of the Siamese structure and flow alignment module, and how these contribute to addressing the identified limitations.

Methods: In the Prediction Module subsection, clarify how the multi-branch encoding-decoding structure network addresses the limited receptive field issue and indistinct change boundaries. Provide a clear explanation of how the network architecture enables the extraction of deep and multi-scale feature information.

 

Results: The abstract mentions "indistinct detection boundaries, poor detection of small targets, and increased occurrences of pseudo changes," but the result section of the manuscript does not elaborate on these challenges. Consider providing a brief explanation of these challenges to help readers understand the specific problems that the proposed algorithm addresses.

Author Response

Thank you very much for your time involved in reviewing the manuscript and your very encouraging comments on the merits. We also appreciate your clear and detailed feedback and hope that the explanation has fully addressed all of your concerns. Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report (New Reviewer)

In this paper, the authors propose a novel change detection algorithm (Siam-FAUnet) that combines Siamese structure and flow alignment module to more accurately extract the changing area and features of the bitemporal remote sensing image. Experimental results show the good performance of the proposed method. However, some weaknesses should be addressed, especially the introduction and experiment.

1) It is suggested that the introduction should be improved. First of all, the research background and significance of this paper are not prominent enough, and why to study this aspect needs further explanation. In addition, the literature analysis in this paper is not comprehensive enough. Sub-pixel level change detection is also an important category, the authors are also suggested to further introduce some other types of change detection methods, such as

[1] Super-Resolution Mapping Based on Spatial-Spectral Correlation for Spectral Imagery [J]. IEEE Transactions on Geoscience and Remote Sensing, 2021, 59(3): 2256-2268.

[2] Subpixel Change Detection Based on Improved Abundance Values for Remote Sensing Images, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2022, 15: 10073 - 10086.

Finally, what is the chapter arrangement? It is suggested to explain at the end of the introduction.

2) The method description needs to be further improved. What is the basis for selecting each module in the network? Can we further prove its effectiveness through the ablation study? In other words, the structure proposed by the author lacks strong interpretability.

3) The experimental data used in the paper is relatively small, and the target of change detection is relatively single. Suggest adding more experimental data. In addition, there is a lack of discussion experiments on the proposed method. Suggest the authors to add some discussion on parameters or operation time.

Author Response

Thank you very much for your time involved in reviewing the manuscript and your very encouraging comments on the merits. We also appreciate your clear and detailed feedback and hope that the explanation has fully addressed all of your concerns. Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report (Previous Reviewer 1)

REPORT REVIEWER

 

The manuscript titled "High resolution remote sensing image change detection method based on improved Siamese U-Net" has been thoroughly revised in its entirety. All sections now exhibit a smooth readability and a decent scientific validity. In particular, improvements have been made to the deficient sections that were identified in the previous review. With that said, I have no further comments regarding the manuscript, which I consider suitable for publication.

REPORT REVIEWER

 

The manuscript titled "High resolution remote sensing image change detection method based on improved Siamese U-Net" has been thoroughly revised in its entirety. All sections now exhibit a smooth readability and a decent scientific validity. In particular, improvements have been made to the deficient sections that were identified in the previous review. With that said, I have no further comments regarding the manuscript, which I consider suitable for publication.

Reviewer 3 Report (New Reviewer)

Thank you for the authors' reply. I don't have any other questions.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

General Comments

 

The manuscript was modified in several parts to achieve greater scientific interest. All comments and suggestions from the previous reviewer were taken into account. However, there are some things that need further modification. At the end of the introduction section, it is important to add a complete and clear aim of the manuscript. In this paper, there is an important error in the discussion section. Usually these are used to compare and comment on the results obtained in this manuscript with those shown in others. Indeed, authors should discuss the results and how they can be interpreted from the perspective of previous studies and working hypotheses. The results and their implications should be discussed in the broadest possible context and the limitations of the work should be highlighted. This comparison is missing in this paper, which is only a focus of the results. In fact, Tab 4 and 5 should not be here but in the results section. I suggest a major revision of the results and discussion sections.

The English language has been well used and only requires minor modifications. However, the simplification of some sentences might be welcome. 

Reviewer 2 Report

First and most important: this submission is not even a finished paper yet! From the draft, I can still see a lot of underscores, strokes,  grammar errors, typo errors, etc. All of these things make the draft very confusing to be understood. Not to mention the errors in almost all figures, like mussy fonts and blocks, overboard content, big stroke, etc.  Due to this, I cannot understand almost all details in this paper. 

See above.