Background Reconstruction via 3D-Transformer Network for Hyperspectral Anomaly Detection

Round 1

Reviewer 1 Report

This manuscript proposes a 3D-Transformer network for hyperspectral anomaly detection. The experimental results on both synthetic and real hyperspectral data sets show that the proposed method is superior to several model-based and AE-based anomaly detectors. This manuscript is well organized, and its content is interesting and innovative especially in the 3DTR network and the patch-generating method. However, some issues still need further explanation and supplement as follows:

1)      The motivation of this manuscript, i.e. the TR module is able to characterize the similarity among pixels, needs to be further discussed and explained.

2)      The colors of each anomaly detector in the ROC curves need to be improved to facilitate the visual observation.

3)      The intervals in Eq. (5) are too large.

4)      The pixels in Fig. 2 seem irregular, and improvements are suggested.

5)      The sizes of legends in Figs. 9, 12 and 15 are too small to read clearly.

6)      The Q, K and V appearing in Eq. (1) are not specified, which is worth modifying and complementing.

 

Good

Author Response

Dear the Reviewer 1,

We would like to express our sincere gratitude to the two anonymous reviewers for their time and suggestions. We found the two anonymous reviewers’ comments very useful in the improvement of this paper. All the comments have been seriously considered and carefully addressed in the revised manuscript.

Further, we will explain the answers to the comments of the Reviewer 1 one by one in the attached file. The corrected and modified parts in the revised version of the manuscript are marked in BLUE color, in order to facilitate reading.

At last but not least, we would like to take this opportunity to thank the Reviewer 1 again for the insightful comments and valuable suggestions, which greatly helped us to improve the technical quality and the presentation of this manuscript.

Sincerely Yours,

Ziyu Wu, and Bin Wang

Sept. 11, 2023

Author Response File: Author Response.pdf

Reviewer 2 Report

06/09/2023

Dear authors,

In the manuscript Background Reconstruction via 3D-Transformer Network for Hyperspectral Anomaly Detection you proposed a background reconstruction framework via 3D-transformer (3DTR) network for the anomaly detection in HSIs. The proposed network utilizes the transformer (TR) module to handle the reconstruction process of each background pixel by its neighbors, allowing for effective characterization of spatial correlations among pixels. By connecting the traditional spatial TR and the presented spectral TR in series, namely the proposed 3DTR network, the background component of a hyperspectral imagery can be reconstructed effectively.

General comments

The study is interesting and experimental results indicate the possibility of wider application of the proposed method in hyperspectral anomaly detection.

The Abstract is too long and does not agree with the instructions to the authors. Yours has 255 words and the instructions say it should not be more than 200 words. The abstract should contain only the basic facts and main results, and all other details should be given in the manuscript.

The Introduction is also too long, especially due to the fact that it is followed by the title Related Works. It should be written more concisely and precisely. This is especially true in the second paragraph where you write about the problems with spectral curves that are mixed, and in the previous paragraph you wrote: ‘Moreover, the spectral curves of different materials are completely different, …’.

Such manuscripts should be written in the third person. Change this throughout the text.

The method is well presented, as well as the experimental results.

I suggest that you announce and describe the materials you used in your research earlier, in the Method and Materials, such as in the instructions to the authors for writing the manuscript. In this way, the results will stand out better, they will not be confused with the description of the images.

Conclusion is too general. In the Conclusion, you should interpret all results (specifically, results) and highlight specific results that are better than the results of existing methods.

Specific comments (are in the manuscript)

1.       Page 1, Abstract - The abstract is too long and does not agree with the instructions to the authors. The abstract should contain only the basic facts and main results, and all other details should be given in the manuscript.

2.       Page 1, Abstract, sentence: ‘The experimental results on both synthetic and real hyperspectral data sets demonstrate the effectiveness and superiority of the proposed method in comparison with several traditional and state-of-the-art (including model-based and AE-based) anomaly detectors.’ - The sentence is too general. Briefly state the specific main results (nominally) of your method.

3.       Page 1, Introduction, sentence: ‘Moreover, the spectral curves of different materials are completely different, …’. Um, is this really so? If you could prove it, every classification result would be ideal.

4.       Page 4, ‘2) In view …’ - Such manuscripts should be written in the third person. Change this throughout the text.

 

Best regards

Comments for author File: Comments.pdf

Author Response

Dear the Reviewer 2,

We would like to express our sincere gratitude to the two anonymous reviewers for their time and suggestions. We found the two anonymous reviewers’ comments very useful in the improvement of this paper. All the comments have been seriously considered and carefully addressed in the revised manuscript.

Further, we will explain the answers to the comments of the Reviewer 2 one by one in the attached file. The corrected and modified parts in the revised version of the manuscript are marked in RED color, in order to facilitate reading.

At last but not least, we would like to take this opportunity to thank the Reviewer 2 again for the insightful comments and valuable suggestions, which greatly helped us to improve the technical quality and the presentation of this manuscript.

Sincerely Yours,

Ziyu Wu and Bin Wang

Sept. 11, 2023

Author Response File: Author Response.pdf