Direction-of-Arrival Estimation over Sea Surface from Radar Scattering Based on Convolutional Neural Network
Round 1
Reviewer 1 Report
REVIEW
Article titled “Direction-of-Arrival Estimation Over Sea Surface From Radar 2 Scattering Based on Convolutional Neural Network”
Remote Sensing no. 1241970
List of authors
Xiuyi Zhao, Ying Yang, Kun-Shan Chen
- In this paper the Authors proposed a novel Direction-Of-Arrival (DOA) estimation approach for SAR systems using the simulated radar measurement of the sea surface at different operating frequencies and wind speeds. This article's forward model is an advanced integral equation model (AIEM) to calculate the electromagnetic scattered from the sea surface.
To solve the DOA estimation problem, the Authors introduce a convolutional neural network (CNN) framework to estimate the transmitter's incident angle and incident azimuth angle. Results demonstrate that the CNN can achieve a good performance in DOA estimation at a wide range of frequencies and sea wind speeds.
- The Authors in Introduction made extensive literature review dealing with the most important aspects about: eestimating the angles of arrival of multiple plane waves, multiple source location, direction of arrival estimation using ANN, parametric localization of distributed sources, effects of local scattering on direction of arrival estimation, and many others.
I am forced to draw attention to a very important substantive problem, because in my opinion, in the Introduction, the Authors should also revision of world bibliography concerning the most important aspects for synthetic aperture radar (SAR) in the aspect of Differential SAR Interferometry technology (DInSAR) and acquisition mechanism of large aperture antennas and frequency bands vs. SAR configurations.
- On the other hand, it is well known that the process of phase unwrapping in SAR technology constrained minimization problem for many well-known algorithms, which are used. Also, a very important problem is acquisition mechanism of large aperture antennas and adaptive forming the beam pattern of antenna in the aspect of the sea surface at different operating frequencies.
For this reason, considering possibilities of SAR technologies, the Introduction of this article should be modified, and some articles for example: “Optimizing the Minimum Cost Flow Algorithm for the Phase Unwrapping Process in SAR Radar”, “Adaptive Forming the Beam Pattern of Microstrip Antenna with the Use of an Artificial Neural Network” and “On the Capabilities of the Deceptive targets generation simulation against multichannel SAR” - should be added to the References.
- In Section 2 entitled “Sea Surface Scattering Datasets Generated by AIEM” the Authors gives the relevant background on the random rough surface scattering models, especially the AIEM model, and illustrates the sensitivity of received signal to influential parameters. It seems to be correct.
Also, I have no comments about the Bistatic Scattering Model in the aspect of mathematical formulas.
- In Section 2.1 (line: 99-101) the Authors wrote: “…AIEM model is an analytical model based on the integral equation method (IEM). Both IEM and AIEM models have been used on sea surface microwave scattering and are in excellent agreement with radar measurements.
My question is as follows. How was the “analytical” model adopted by the Authors verified in reality?
- The level of dispersion/scattering is directly dependent on the frequency of the signal. How did the Authors of this article investigate this issue?
The research was carried out only for a few selected frequencies (see: Table 1).
The above requires a very precise comment from the Authors of this article.
- Another concern is the performance comparison with other algorithms. I suggest the Authors to also perform more experiments with more datasets, where they can also demonstrate the applicability of the proposed method on real data (not only, as now), because that might show some performance benefits of the proposed method/algorithm.
- Some information about the time complexity in the aspect of proposed method - should be also added.
The Authors addressed a problem which is relevant and appealing for this Journal. However, I cannot recommend the current manuscript for publication unless the current version is corrected. After providing the amendments to the article, the work ought to be reviewed once again.
Comments for author File: 
Comments.pdf
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 2 Report
The authors describe a simulation based study concerning the estimation of the direction of arrival of radar waves affected by sea scattering by means of convolutional neural networks. The manuscript is well written from the point of view of the language: only one minor typographical error was found on line 27 ("an fundamental" in place of "a fundamental").
The introduction is also of good quality: a satisfying number of reference to state-of-the-art papers is present and the description provides a good background to the problem of radar DOA estimation as a whole. Section 2 is very clear: the mathematical model is sound, and the notation does not generate confusion into the reader. Sections 3.1 and 3.2 are also fine.
Section 3.3 is definitely too short and brief and must be extended. Considering the fact the CNN is mentioned in the title of the manuscript, the amount of space devoted to the description of the chosen CNN structure and working principle is not enough. In figure 4 numbers are given for each stage of the CNN: what are the mathematical criteria on which these numbers were preferred in place of others? Why using 6 layers instead of 5 or 7? Are there computational time/accuracy trade-off that led to this structure? How the accuracy of the final prediction is related to number of elements in the regression layer? Would using 512 or 2048 have resulted in better or worse results?
Similar considerations can be made for section 4. No actual comparison to the state-of-the-art is made by the authors, so it is not possible to evaluate how the proposed algorithm performs in relative terms. How does an RMSE error of 1 degree compare with other algorithm in literature? Is it enough to locate an object at a give distance? What about the azimuth error accuracy? What about the computational processing time? Is it enough to perform real-time target tracking?
In absence of these comparisons it is strong opinion of the reviewer that the conclusion the authors drawn that accuracy is satisfying is non supported by facts. The quality of the work must be increased with comparisons showing that the proposed results are in-line or better than the state of the art before considering publishing this work.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
Dear Authors,
the aim of this paper is more or less clear. What is necessary to improve is some technical passages related, for example, to this article: https://www.mdpi.com/2072-4292/11/5/596/htm It would be better to explain the differences between these approaches.
Then, there are some parts not clear from the point of view of the language, please review the whole paper using an English mother-tongue.
Due to the above, my final recommendation is minor review.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
REVIEW_2
Article titled “Direction-of-Arrival Estimation Over Sea Surface From Radar Scattering Based on Convolutional Neural Network”
Remote Sensing no. 1241970
List of authors
Xiuyi Zhao, Ying Yang, Kun-Shan Chen
The article MDPI no. 1241970 Remote Sensing titled “Direction-of-Arrival Estimation Over Sea Surface From Radar Scattering Based on Convolutional Neural Network” has been carefully modified and well revised.
The present version of the article includes all remarks found in the reviews.
In this way, present version of this article may be finally accepted for publication in MDPI Remote Sensing.
Comments for author File: Comments.pdf
Reviewer 2 Report
The authors describe a simulation based study concerning the estimation of the direction of arrival of radar waves affected by sea scattering by means of convolutional neural networks. The revised manuscript is well written from the point of view of the language and has been largely improved by the authors, addressing all the points raised by the reviewer.
The added information in section 3.3 has been particularly appreciated, as well as the highly recommended comparison with state-of-the-art in section 5.2. The added references also improved the overall quality of the work and help correctly benchmark the accuracy level of the proposed algorithm.
According to these considerations, the manuscript can be accepted for publication in the present form.
