SAR Image Reconstruction of Vehicle Targets Based on Tensor Decomposition
Round 1
Reviewer 1 Report
This manuscript is related with image reconstruction.
Some suggestions are as follows:
1. In the abstract is "There is strong correlation and redundancy between adjacent azimuth images of SAR targets". Please shortly explain this.
2. "This shows that the method proposed in this paper can support target recognition under the condition of sparse observation to a certain extent."
Please explain this sentence.
3. "The matrix filling process is represented by the example shown in the figure below."
I suggest ".... in the Figure 1."
4. Where is the reference to Figure 3 in the text?
5. What is presented in Figure 6? There is no information in manuscript.
6. There is no information in the text about Figure 8.
Author Response
- In the abstract is "There is strong correlation and redundancy between adjacent azimuth images of SAR targets". Please shortly explain this.
Response：
Thank you for your comments.
The MSTAR data set are used our the experiment. Many vehicles targets, such as T72, BMP2 and BTR70, have azimuth images with an interval of 1 degree. Through research and analysis, the SAR images have high similarity and strong correlation within the adjacent azimuth angle from 1 to 5 degrees. For template based target recognition, too many sample data sets will also lead to the increase of calculation time and the decrease of efficiency, which is the reason for its redundancy.
- "This shows that the method proposed in this paper can support target recognition under the condition of sparse observation to a certain extent."
Response：
Please explain this sentence.
Thank you for your comments.
In our paper：‘When the azimuth images are separated by 10°, the reconstructed image still has a high feature similarity with the original image. This shows that the method proposed in this paper can support target recognition under the condition of sparse observation to a certain extent.’
Even if the azimuth angle of the missing vehicle target SAR image is 10 degrees apart from the azimuth angle of the existing data, it can still be reconstructed by this method to maintain high feature similarity. Under the condition of sparse observation, there is a large azimuth interval between the acquired observation data and the existing template data. Through reconstruction, the similarity between the same target recognition template and the observation data can be ensured to solve the recognition problem.
- "The matrix filling process is represented by the example shown in the figure below."
I suggest ".... in the Figure 1."
Response：
Thank you for your comments. As you said, it should be in Figure 1. We have revised the relevant text and marked it in blue in the revised draft.
- Where is the reference to Figure 3 in the text?
Response：
Thank you for your comments. Step (II)-(IV) is shown in Figure 3. We have revised the relevant text and marked it in blue in the revised draft.
- What is presented in Figure 6? There is no information in manuscript.
Response：
Thank you for your comments.
Figure 6 is a description of MSTAR data. We have adjusted Figure 6 to the Section IV. The experimental data are introduced at the beginning of the Section IV.
- There is no information in the text about Figure 8.
Response：
Thank you for your comments. Figure 8 is a partial enlarged view of the accuracy of Figure 7, which is explained in the revised version and marked in blue.
Finally, thank you for your review. Your comments will help improve the quality of this article.
Reviewer 2 Report
This review paper is considered an interesting and valuable paper. And a probabilistic approach to this paper seems possible. Therefore, I recommend the following papers to the authors, and it would be better if authors could provide a brief comment for the researchers.
Kim, T.; Kim, D. S.; Lee, H.; Park, S.-H. Dimorphic properties of Bernoulli random variable. Filomat 36 (2022), no. 5, 1711–1717.
Kim, T.; Kim, D. S. Degenerate zero-truncated Poisson random variables. Russ. J. Math. Phys. 28 (2021), no. 1, 66–72.
Author Response
Thank you for your comments. We cite the relevant literature and give a brief introduction in the introduction.
Reviewer 3 Report
Dear Authors,
You have considered an interesting research subject. However, below are some notes you need to consider to improve the article.
1- The obtained results need to be better reflected by values or percentages in the abstract.
2- Some symbols and abbreviations are mentioned in the mathematical formulation without a prior definition.
3- Please write a short paragraph at the end of the introduction section to show how the paper is organized.
4- Please give more explanation about Figure 1.
5- Based on what the k1 and k2 parameters were selected.
6- Please give more details about Figure 4.
7- More details about the designed convolution neural network are needed.
8- This sentence, “The main parameters in the 196 non-negative Tucker decomposition are” in line 196 on page 6, is incomplete; please check it.
9- In the conclusion section, please refer to the future prospects of this work.
10- There are some typos and grammatical errors in the manuscript; thus, it is strongly recommended that the whole work be proofread carefully.
Author Response
1. The obtained results need to be better reflected by values or percentages in the abstract.
Response：
Thank you for your comments.
In the abstract, we have described the numerical indicators of the relevant results marked in blue.
2. Some symbols and abbreviations are mentioned in the mathematical formulation without a prior definition.
Response：
Thank you for your comments.
We have carefully corrected the errors and omissions in the relevant formula symbols and variable definitions.
3. Please write a short paragraph at the end of the introduction section to show how the paper is organized.
Response：
Thank you for your comments.
At the end of the Section I, we added the full-text organization description.
4. Please give more explanation about Figure 1.
Response：
Thank you for your comments.
We have added the specific description of Figure 1 and marked it in blue.
5. Based on what the k1 and k2 parameters were selected.
Response：
Thank you for your comments.
In the reference [18], the author designed the calculation formula of SSIM, defined that k1 is a positive number far less than 1, and set K1 = 0.01 and K2 = 0.03 in their relevant experiments. This paper uses these parameters for reference.
6. Please give more details about Figure 4.
Response：
Thank you for your comments.
Figure 4 is a schematic diagram of SAR image feature extraction using convolutional neural network. We also added a description of the convolutional neural network settings.
7. More details about the designed convolution neural network are needed.
Response：
Thank you for your comments.
Figure 4 is a schematic diagram of SAR image feature extraction using convolutional neural network. We also added a description of the convolutional neural network settings marked in blue.
8. This sentence, “The main parameters in the 196 non-negative Tucker decomposition are” in line 196 on page 6, is incomplete; please check it.
Response：
Thank you for your comments.
We have carefully corrected the errors and omissions.
9. In the conclusion section, please refer to the future prospects of this work.
Response：
Thank you for your comments.
In the conclusion, we added arrangements for future work.We have carefully corrected the errors and omissions marked in blue as following.
The future research work mainly considers the following two parts. First, the re-construction and verification of SAR image data of other vehicle targets are added. On the other hand, the reconstructed data will be used for vehicle target detection and recognition in SAR images to measure its practicability and robustness.
10. There are some typos and grammatical errors in the manuscript; thus, it is strongly recommended that the whole work be proofread carefully.
Thank you for your comments. We have carefully corrected the errors and omissions.
Finally, thank you for your careful review. Your comments will help us improve the quality of this article.
Round 2
Reviewer 3 Report
Thanks for considering all the given comments
