Deep Convolutional Denoising Autoencoders with Network Structure Optimization for the High-Fidelity Attenuation of Random GPR Noise

Round 1

Reviewer 1 Report

Dear Authors,

Thank you for the paper and interesting results. This paper presents a study of ovel network structure for convolutional denoising autoencoders (CDAEs). Results presents a very interesting radar images in different condition, and the clarity of the image due to enhancing the SNR.  I further, propose the following comments on the paper;

Comments:

1- Introduction is divided into two parts first general GPR applications, which is not very well addressed, I would suggest to re-consider this part and explain more about the GPR applications, and why your approach is important and in which applications could have significant results for example in case of water content, or saturated clay? I propose to investigate a bit more in state-of-the art? I propose below articles.

Solla, M.; Lagüela, S.; Fernández, N.; Garrido, I. Assessing Rebar Corrosion through the Combination of Nondestructive GPR and IRT Methodologies. Remote Sens. 2019, 11, 1705. https://doi.org/10.3390/rs11141705

Rasol, M. A., Pérez-Gracia, V., Fernandes, F. M., Pais, J. C., Santos-Assunçao, S., Santos, C., & Sossa, V. (2020). GPR laboratory tests and numerical models to characterize cracks in cement concrete specimens, exemplifying damage in rigid pavement. Measurement, 158, 107662. https://doi.org/10.1016/J.MEASUREMENT.2020.107662

Li, C. Le Bastard, Y. Wang, G. Wei, B. Ma and M. Sun, "Enhanced GPR Signal for Layered Media Time-Delay Estimation in Low-SNR Scenario," in IEEE Geoscience and Remote Sensing Letters, vol. 13, no. 3, pp. 299-303, March 2016, doi: 10.1109/LGRS.2015.2502662.

 

2- In the second part of introduction, I would suggest please to insert a table and summarise the idea of developing considering other studies, and how SNR is enhanced by different approaches? It would be very interesting and easy to read not so confused. 

3.Figure 1, need to be more well addressed as it is the main idea of the work,  the explain better the idea of each depth in layers, if it is possible.

4. In section 5.2, please described which is the raw data and which processed with your approach, will be good to compare with usual post-processing approaches and compare three cases. I refer to figure 9.

5- I would also suggest considering the limitation of the approach and this could affect emitting some important information in some cases and explain the complexity of each approach.

6- Conclusion is not well addressed, please consider validation methodologies of results with the conclusion.

7- I find it very interesting if you can make a comparison table between each approach; raw data, CDAEs, AD-CDAEs and AD-CDAEs-ResNet considering several GPR signal parameters which can be affected by this analysis.

Please consider re-review according to the above-mentioned comments, I suggest a major revision of the paper.

 

 

Author Response

Please see the attachment, thanks.

Author Response File: Author Response.pdf

Reviewer 2 Report

The work introduced an autoencoder-based strategy for the noising problem in GPR survey. I recommend improving the paper in two aspects. 
1. The content is not convincing enough. The authors mentioned that the autoencoder-based strategy has the advantage of self-adjust according to the characteristics of various signals compared to the conventional methods. However, the comparison between the proposed CDAEsNSO and the conventional methods was not found in the experiments.
2. The expression was too verbose. For instance, readers could not capture the authors’ intention until the 3rd page. It's hard to get the logic quickly and clearly. Besides, the authors’ language needs polishing. Seems that the paper was translated directly from the native language.

Generally, impressive work.

Author Response

Please see the attachment, thanks.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Dear Authors,

Thank you for addressing all the comments. I congratulate your interesting work and contribution to the scientific community. I accept the paper for publication as in this form of the manuscript. 

Looking forward to hearing back more works from your team.