Attention-Based Matching Approach for Heterogeneous Remote Sensing Images

Round 1

Reviewer 1 Report

The manuscript deals with the problem of matching remotely sensed images taken from different sensors with different characteristics (spatial/spectral resolution, technology). Authors propose a new method, CSAM, mainly based on cross-attention and self-attention methods. As a by-product, authors develop a data set of 700k heterogeneous images for the analysis. Authors compare their methods with some state-of-art competitors.

Paper is very well written, with a rich literature review. The proposed methodology is innovative. Results are comparable and in many cases better than best competitors as far as quantitative and qualitative indicators are concerned; in addition the computational time is better than the methodologies considered for comparison.

Therefore I recommend publication of the manuscript in the Journal

Minor typos:

l. 258-259: missing formula
l. 293: maybe "use" instead of "uses"
l. 301: maybe remove "but"
Fig. 15: numbering of subfigures should be (e) (f) instead of (d) (e)

Author Response

Thank you for your letter and the reviewers’comments on our manuscript entitled "Attention-Based Matching Approach for Heterogeneous Remote Sensing Images" (ID：remotesensing-1991277). Those comments are very helpful for revising and improving our paper, as well as the important guiding significance to other research. We have studied the comments carefully and made corrections which we hope meet with approval. The main corrections are in the manuscript and the responds to the reviewers’ comments are in the attachment (the replies are in red ).

Author Response File: Author Response.docx

Reviewer 2 Report

Small changes needed - see attached - overall a good paper

Comments for author File: Comments.pdf

Author Response

Thank you for your letter and the reviewers’comments on our manuscript entitled "Attention-Based Matching Approach for Heterogeneous Remote Sensing Images" (ID：remotesensing-1991277). Those comments are very helpful for revising and improving our paper, as well as the important guiding significance to other research. We have studied the comments carefully and made corrections which we hope meet with approval. The main corrections are in the manuscript and the responds to the reviewers’ comments are in the attachment (the replies are in red ).

Author Response File: Author Response.docx

Reviewer 3 Report

The authors proposed attention-based matching approach for heterogeneous remote sensing images.

However, the presented approach has minor improvement in various experiments and the improved processing time is very small. Also, the test data which used in this study for comparison is very small compared to the training data only 10000 images from 700000. Furthermore, there are conflicts in the results especially for the UAV optical infrared images.

Considering that the topic of the study is interesting, they should improve the introduction section and discuss the advantages and disadvantages of the proposed approach in a separate discussion section. Also, some figures are not clear and should be modified so that the final publication has a good impact.

In general, there structure and many English language errors scattered throughout the manuscript, for which the authors need to account for.  The manuscript should have English proofreading.

Thus, I suggest a reconsidering this manuscript after major review.

Please find below my comments and recommendations.

 

Abstract:

Lines 12-15: Please rephrase this paragraph.

Line 16: Please add the CSAM abbreviation here.

 

1-    Introduction:

Lines 39-41: Please rephrase this paragraph.

Line 52: Please remove the (see, e.g.) sentence and the same through the manuscript.

Lines 67-68: Please add a reference to prove this hypothesis.

Lines 88-93: Please rephrase this paragraph.

Lines 97-100: Please rephrase this paragraph.

Lines 106-108: This sentence should be at the end of the introduction section.

Line 109: Please add a reference after et al.

Line 111: Please remove the repeated word.

Lines 121-155: These paragraphs should be moved to the methods section.

 

2-    Methodology.

Line 163: Please clarify the Q (n) symbols.

Lines 232-235: the test data is very small compared to the training data. I suggest split the dataset to 60% training data, 10% validation, and 30% test data.

Line 258: Please clarify be what?

 

3-    Experiments.

Lines 313-315: Please clarify if the hardware requirements are important for these experiments and whether it needs a GPU card or not.

Lines 317-322: Please rephrase this paragraph and it’s better to illustrate this method in numbers.

Line 377: Please add the threshold numbers to figure 10 a. Also, the time difference between the tested layers can be neglected.

Line 390: Table 1 both the CFOG and RIFT methods have better precision than the proposed approach. Also, the time difference between the tested layers can be neglected.

Lines 398-402: Please rephrase this paragraph.

Lines 408-411: This paragraph should be addressed before the figures not after.

Line 441: Table 3 the RIFT + NNDR + FSC methods have better precision than the proposed approach.

Lines 469-471: This paragraph conflict with the results in Table 3.

 

4-    Conclusions: 

Line 487: Please remove the repeated word.

Line 492: Please clarify this percentage is accuracy or precision.

 

Author Response

Thank you for your letter and the reviewers’comments on our manuscript entitled "Attention-Based Matching Approach for Heterogeneous Remote Sensing Images" (ID：remotesensing-1991277). Those comments are very helpful for revising and improving our paper, as well as the important guiding significance to other research. We have studied the comments carefully and made corrections which we hope meet with approval. The main corrections are in the manuscript and the responds to the reviewers’ comments are in the attachment (the replies are in red ).

Author Response File: Author Response.docx

Reviewer 4 Report

The manuscript proposed a feature point matching method, namely CSAM, based on attention. The proposed method achieved robust cross-modal image matching on heterogeneous remote sensing images. The manuscript is very well written and presented. Extensive experiments on the proposed dataset in this manuscript validate the effectiveness of CSAM. I would recommend a minor revision to address general and specific comments reported below.
1.In the line 35-36, Some literature about cross-modal image retrieval are omitted in the manuscript, e.g., [1]. Please cite the relevant literature.
[1]Y. Fang, P. Li, J. Zhang and P. Ren, "Cohesion Intensive Hash Code Book Coconstruction for Efficiently Localizing Sketch Depicted Scenes," in IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1-16, 2022, Art no. 5629016, doi: 10.1109/TGRS.2021.3132296.
2. Some failure matching cases should also be shown. The relevant discussion can be added in the conclusion.

Author Response

Thank you for your letter and the reviewers’comments on our manuscript entitled "Attention-Based Matching Approach for Heterogeneous Remote Sensing Images" (ID：remotesensing-1991277). Those comments are very helpful for revising and improving our paper, as well as the important guiding significance to other research. We have studied the comments carefully and made corrections which we hope meet with approval. The main corrections are in the manuscript and the responds to the reviewers’ comments are in the attachment (the replies are in red ).

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

The authors proposed attention-based matching approach for heterogeneous remote sensing images.

However, the authors only perform the minor comments, did not improve the manuscript, and the major comments still not revised.

Once again, the presented approach has minor improvement in various experiments and the improved processing time is very small. Also, the test data which used in this study for comparison is very small compared to the training data only 10000 images from 700000. Furthermore, there are conflicts in the results especially for the UAV optical infrared images.

Moreover, they should improve the introduction section and discuss the advantages and disadvantages of the proposed approach in a separate discussion section. Finally, more recent references should be added to support the authors hypothesis.

In general, there are style, structure and other English language errors scattered throughout the manuscript, for which the authors need to account for. The English language should be reviewed by native English-speaking editors. Also, the article is very complex and hard to follow, and the introduction and the conclusion sections should be revised.

Thus, I suggest reconsidering this manuscript after performing the abovementioned comments.

 

Author Response

Thank you again for your comments on our manuscript entitled "Attention-Based Matching Approach for Heterogeneous Remote Sensing Images" (ID：remotesensing-1991277). Those comments are very helpful for revising and improving our paper. We have studied the comments carefully and made the appropriate corrections, which we hope meet your approval standards. The main corrections are in the manuscript, and the responses to the reviewers’ comments are presented in the attachment (the replies are in red ).

Author Response File: Author Response.docx