Structure Label Matrix Completion for PolSAR Image Classification

Round 1

Reviewer 1 Report

I agree with the present paper.

Author Response

We thank the Reviewer for the positive comments about our work and manuscript.

Reviewer 2 Report

see attached file.

Comments for author File: Comments.pdf

Author Response

We thank the Reviewer for the comments about our work and manuscript. Below, we address every comment carefully and explain the corresponding changes in the manuscript.

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors in this manuscript (remotesensing-665514) proposed a matrix completion framework instead of the classification task for Polarimetry SAR imagery. The uniform under-sampling PolSAR image is then introduced to determine the known labels. They applied the proposed method on the Flevoland PolSAR dataset from NASA/JPL AIRSAR and Oberpfaffenhofen from ESAR Airborne. The manuscript needs English proof-reading because there are some grammatical issues. Overall, this is a properly organized manuscript. I also do have some comments as follows:

Abstract: The abstract in the current version is not informative. The abstract should be short enough for readers to scan quickly and long enough to give them enough information to decide to read the article. The abstract should be rewritten another time since, in the current version, it is more similar to the method summary. In the abstract you need to indicate the type of information found in the paper; explains the purpose, objective, and methods of the paper as well as your contribution and show the improvements that were obtained compared to other methods.

Good introduction Page 2, Line 35: Add Fully Convolutional Neural Network to machine learning examples and provide the following citations for readers: Lin, H., Shi, Z. and Zou, Z., 2017. Fully convolutional network with task partitioning for inshore ship detection in optical remote sensing images. IEEE Geoscience and Remote Sensing Letters, 14(10), pp.1665-1669. Mohammadimanesh, F., Salehi, B., Mahdianpari, M., Gill, E. and Molinier, M., 2019. A new fully convolutional neural network for semantic segmentation of polarimetric SAR imagery in complex land cover ecosystem. ISPRS journal of photogrammetry and remote sensing, 151, pp.223-236.

 

Page 8, Remote equation 10 and add a reference instead. I would suggest adding a machine learning classifier such as RF and SVM to compare the results. Conclusion: The authors should add a result summary that illustrates the performance of the proposed method in comparison with other methods in conclusion.

 

 

Author Response

We want to thank the reviewer and his/her positive comment on this paper. Below, we address every comment carefully and explain the corresponding changes in the manuscript.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

My previous comments are not completely addressed. The major problems include incomplete experimental results and inappropriate experiments design. I still have the following questions.

If the experimental results are reported in the average of five runs, the standard deviation should be reported as well. Reference method CLSL is still not well explained. A simple description to this approach is required. At least its general focus and application should be provided. Response 5: 1) Kappa coefficient is not always consistent with overall accuracy (imbalanced data), so it is necessary to report it for all the methods. 2) If it is unrealistic to run SAG method in the ESAR data, how did authors obtain the results in the table? I agree with that normally classifiers with more training samples yield improved performance, but it is still not reasonable to compare with other methods by using different amount of labelled samples. The experiments have to be reconducted in a fair way. Different classifiers should utilize the same labelled training data. Some classification maps (SAG, SRDNN) are still missing. The caption in Figure 9 is wrong.

Author Response

 We thank the anonymous Reviewers for their efforts and feedback. Below, we address every comment carefully and explain the corresponding changes in the manuscript.

Author Response File: Author Response.pdf

Round 3

Reviewer 2 Report

In the response to the Comment 1, the English writing is very bad. Also there are many grammar mistakes.  In case that some results are obtained from the reference paper, it is better to mention this in the paper.

Author Response

 We thank the anonymous Reviewers for their efforts and feedback. Below, we address every comment carefully and explain the corresponding changes in the manuscript in the attachment.

Author Response File: Author Response.pdf