Attention-Guided Siamese Fusion Network for Change Detection of Remote Sensing Images

Round 1

Reviewer 1 Report

The presented mansucript reflects a considerable efforts and tackles an important field within remote sensing. However, after a careful investigation of the manuscript, and under its current form, the manuscript is unsuitable for publication. It requires very significant efforts to be reconsidered. The following comments are povided for support.

General comments

The Authors do not provide any elements that clearly explain their objectives, contribution to the field and the novelty of the manuscript.

The manuscript is very badly written and the poor use of English makes it very unpleasant to read. 

The number of Figures is very excessive and it can be confusing to readers to keep track of what is presented and meant.

Specific comments

Abstract

a. Line 11 : The article starts with a very badly written sentence. In addition, how come is et al. inserted in the abstract after the word monitoring?
b. Line 14: is an indispensable procedure not "is the"
c. Line 17: Weather et al ?!
d. In addition to the poor use of English, the objectives aren't clearly presented nor are any results displayed or explained.
e. The abstract should be completely rewritten.

1. Introduction

a. No clear objectives are listed. The innovative aspect is very badly described and not clear.
b. Some elements, such as lines 94-97 are clearly not the Authors' findings and should be supported by relevant literature.
c. Lines 130-131: The introduction does not report findings, please remove this sentence

2. The proposed method
a. This section severly lacks any reference to other works and citations. Even if the Authors elaborated this method, it should be well supported by literature. It is very weird to see more than 4 pages with only 2 citations. The Authors also do not indicate what is the novelty of this approach and what gaps in the domain is it filling.

3. Results

a. This section is very rich and the strong point is the comparison with other models. The Authors elaborated well their findings, and the latter are well tested for statistical significance. 
b. Line 426: Mind the Fig. clearly refer to the concerned Figure
c. With that said, I am afraid that 20 figures is a very excessive number. The Authors should consider merging some or even removing others. The way the results section is presented resembles a report more than a scientific paper
d. The same could be said for Tables

4. The conclusion section is way too short, unexplicative and repeats what has been said previously. It should be rewritten to overarch accross the objectives, summarize the methodology, reveal innovations, display brief results, briefly discussing them and proposing future works

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Introduction 

Theoretical issues are useful and help very much to argue the formulation of the paper objectives. I think the section can be a little shorter and better organized between paragraphs, from simple to complex methodologies. Change detection analyses from literature can be fitted also with specific examples from real problems like urban area, forestry, agriculture, mining etc. They are mentioned but the methodologies could be better adapted to some types of approaches and types of remote sensing data. 

Methodology section

Theoretical issues are necessary and need more explanation for the normal reader interested in searching for a more objective remote sensing imagery change detection focused analysis. Each scheme needs an explanation of the steps better fitted with the mathematical transformation of multitemporal image raster data. 

The basic module is an interesting tool but it needs more accessible adapted explanation.

There are so many acronyms. Most of them are explained but a table with them can be helpful for the normal reader like an Earth scientist, for example.

A synthetic table of the featured flow charts/transformation modules could help the reader to extract the essential aspects.

Results section (experiments)

Figures 7-12 are image samples not enough explained in terms of features and change detection analysis objectives. In fact is a collection of experimental sites for which the reader need to know the reason of selection and even the location on a map. A synthetic table can be helpful to explain the tests used data sets. In fact change detection proposed approach need to be tested on a large complementary numer of real life situations.

Technical details needs a more simple explanation for a large category of readers, as remote sensing change detection is an issue of an outstanding interest in literature.

Tables 2-5 are essential for the results and need to be better explained in order to extract the core of the approach.  Accronyms can be explained to the table and not only in the text methodological section.

Section 3.3. The comparison is useful and needs to be presented as a discussion separate section of the manuscript.

Calibration of the current methodology is essential. You need to provide more explanation for tables 6-11 and figures 16-21. It is important to explain the sense of ground truth map and the significance of binary encoded layers produced by change detection analysis. Is it enough to map only the spatial-temporal change and less the magnitude of change.

Some examples of change detection results interpretation can be helpful. What features from original imagery influence the results in terms of accuracy and significance?

Are noise, radiometric calibration of basic imagery, resolution, radiometric resolution or pixel depth or other aspects influencing the results in mapping changes in land use and land cover?

Is it possible to extract the best results in terms of accuracy for different landscape features? For example the approach is better adapted to agricultural landscapes than urban/rural landscapes etc.

Conclusion

It can be interesting to provide a practical example for the application of methodology in a social and economic significant issue. 

The approach is really useful and original but a more detailed interpretation of results is needed. 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

In lights of the revised version, the following comments are listed:

1- Abstract: the use of English is still very bad and insights on results have not been presented.
In the Author's response the following is listed: The grammatical and clerical errors also are checked and ignored as far as possible - what is meant by ignored ?
2- Introduction: same as the abstract, a manuscript with a bad level of English is very problematic, especially for international reviewers. I strongly suggest a revision from a native English speaker as it is very unpleasant to read.
3- A glossary of terms cannot be added within the main text. Define the terms next to their acronyms between brackets or parentheses.
4- The ground truth mad by humans? I assume Authors mean made? If yes, this also is wrong it should be ground truthing performed by ....
5- man-made building? The very poor use of English is a serious flaw. Any buiding is man-made they do not grow naturally!!!
6- Other comments have been well-addressed yet the mansucript still lacks clarity, and ease of reading flow.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript was revised in detail and is easier to be followed and understood. Please check carefully the orthography and the correspondebcw between references and text numbering. It is good you explained the significance of results.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 3

Reviewer 1 Report

Dear Authors,

Congratulations on the changes you have performed, the manuscript is now well presented and readers can understand well the value of your important work.