A Novel Method to Identify the Spaceborne SAR Operating Mode Based on Sidelobe Reconnaissance and Machine Learning

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

 

Comments to the Author:

 

Based on sidelobe reconnaissance and machine learning, a novel method to identify the spaceborne SAR operating mode is proposed in this manuscript. The neural network algorithm is employed to extract the antenna pattern features of the SAR signals and improve computing efficiency. The simulation verifies the feasibility of the proposed identification technique, and the influence of important parameters on identification performance is analyzed. The paper is well organized and contains interesting material that if properly revised can deserve publication.

 

The specific comments are listed:

1. The English language must be carefully checked and improved, such as “The problem of the SAR operating mode identification was transformed, …, to obtain higher computing efficiency than conventional global search algorithms”. It is suggested to divide a long sentence into several shorter sentences.

 

2. Another obvious problem with this paper is lack of sufficient explanation of the simulation results. You need to explain your simulation results in detail and why you got such results.

 

3. Conclusions needs more in it, as it's more of an afterthought. The authors are suggested to highlight important findings and include afterthought of this work, such as a creative outlook and a future research plan.

 

4. The paper needs a summary of the limitations of this method and the possible coping strategies should be given.

 

5. It is suggested that the authors carefully check the references to ensure that the format and annotation are accurate and standardized.

 

Comments on the Quality of English Language

 Minor editing of English language required

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Major comments:

1.       A deep learning approach was used, but this level of complexity is not motivated. Comparison of results to other methods from literature should be presented. Observing Figure 6 one might even think that machine learning is not needed, since a simple threshold on the derivative can be enough.

2.       Presenting results only in terms of accuracy tables is not convincing, and some examples should be included. For example, in which cases does the method struggle to perform.

3.       The theoretical part should be revised. Equation 8 Is equal to 12 and 9 is equal to 13. So spotlight and Stripmap are the same? Surely not, otherwise the different plots in figure 5 would be strange. Models should be verified.

Minor comments:

1.       Abstract should better explain the goal of operating mode identification. It becomes clear only after reading the article that you are talking about a military application, active jamming, and that you want to intercept SAR signal from the ground and to add fake targets.

2.       Lines 90-94 are repeating lines 85-87.

3.       Equations 3 and 4 are repetitive. You can convert remove the last part of equation 3, and write 4 as P_rf / L_r.

4.       Line 175 SAR and target are given in (ground range, satellite position and z), while jammer in x,y,z? the different coordinate systems should be clarified. Also, usually x is parallel to the orbit direction, so the notation is confusing.

5.       Seems that figure 2 and 3 are identical apart for the swath notation, can be united.

6.       Figure 5: add axes units in m/km, and provide colorscale and units of power (dB?).

7.       Figure 9: define accuracy and cost.

 

8.       Fractional Fourier transform is mentioned in 251 and explained in 310. Explanation should come first. Also, details about implementation (line 305 – 315) should not be in simulations section.

Comments on the Quality of English Language

Overall the language is clear, by some sentences are difficult to understand. For example:

1.       Line 46: “This method has good extensibility, and this paper will take the above modes as an example to carry out experiments. 47 without loss of generality”. Which modes? Spotlight or all (stripmap, scanSAR and spotlight).

2.       Line 81: However, in practice, the jammer based on main-lobe reconnaissance cannot provide overhead warning in advance”: the sentence is not clear. 

3.       Line 183: the sentence is not clear.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This paper proposed a novel identification method for SAR operating mode, where the sidelobe reconnaissance equation and the neural network are employed. In this manuscript, the intercepted signal is modelled in detail. Besides, the experimental results show that this method has good robustness to low-SNR and the uncertainty of the jammer position. The paper is practical for engineering applications and the results are original. I recommend the publication of the paper after the following minor corrections:

1. In abstract: please add more sentences for finding of the results.

2. Correct language errors and improve the narrative and clarity of the text to make the manuscript more readable. For example: The use of compound words is not uniform, such as “main lobe” and “main-lobe”.

3. Figure 2: please add the explanation of  in the caption. This will increase the readability of the figure.

4. Please add the specific references in the first sentence in subsection 3.1 (When conducting sidelobe reconnaissance ...... peak value of received pulses).

5. The formula (11) should be centered.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The reviewed article is much improved in terms of readability. The flow is good, and it is possible to understand the concepts. However, some results and motivations are missing.

Major issues:

1.      It should be clear why a deep learning method is needed, which is not obvious from Figure 6. It seems that just looking at the range of values of the dynamic range should be enough to identify Spotlight mode, for example. It will be useful if some examples of difficult cases are shown (like the one you added in the review report, even if it is not clear why in the spotlight mode there is a local minimum, different from Figure 6).

2.      Confusion matrices should be added to show how the algorithm is handling the different modes. Is it more easy to detect one of the modes rather than others?

3.      In which cases the algorithm fails? Examples should be provided to make the results more clear.

4.      Details about the training/testing sets should be provided. How many signals are used in each?

5.      The comparison with state of the art should be improved. Specifically: Figure 4, The GRU method is not presented elsewhere (or at least the abbreviation ), how does it work? If the results are similar, what is the benefit of your method? Should be commented.

 

Minor issues:

1.      Line 48: “This method” – to which method do you refer?

2.      Figures 2,3: mark Rg

3.      Line 180 “hc is beam center crossing time of the point target;”  which point target? Not clear

4.      Figure 5: usually units are marked in brackets [km]

5.      Provide equations for accuracy and cost.

 

6.      Table 6: in the case of Spotlight SAR, if the position is at the minimum of the antenna pattern, wouldn’t the signal always be almost zero? How could you make a prediction in this case? A confusion matrix is needed to understand how the algorithm performs w.r.t all the modes. 

Author Response

Please see the attachment

Author Response File: Author Response.pdf