An Improved Sea Ice Classification Algorithm with Gaofen-3 Dual-Polarization SAR Data Based on Deep Convolutional Neural Networks

Round 1

Reviewer 1 Report

In this paper, the authors mainly discuss an improved sea ice classification algorithm with gaofen-3 dual polarization SAR data based on deep convolutional neural networks. In general, the work in this paper is well described, and the organization is well organized. However, some minor revisions are still needed before acceptation.

1. As the authors said that the parameters in Table 3 have a significant influence on the training effect. That is to say, the direct setting in Table 3 is not convincing. At this point, the authors should discuss the influence of these parameters on the performance. Then, the optimum parameters can be visually found by readers.
2. In Eq. (6), the reviewer wanders to know how to obtain the probability function p and q.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Dear Authors,

Your work “An Improved Sea Ice Classification Algorithm with Gaofen-3 Dual Polarization SAR Data based on Deep Convolutional Neural Networks” performs a Sea-Ice characterization based on Gaofen-3 Dualpol products using Machine Learning algorithms and suggesting a Multi-Scale MobileNet (MSMN) DCNN as the best performing method, compared to other DCNN.

General comments:

The manuscript is well written and structured. The description of the research methodology is correct, although sometimes, it is difficult to follow the explanations. However, this something common to all the papers including this kind of procedures.

In general, you follow a standard classification methodology, where the novelty is the Multi-Scale MobileNet (MSMN) DCNN. Regarding the results, it is interesting to observe a better classification of Sea Ice types by only single polarization states rather than using dualpol states.

Discussion and conclusions sections are also well presented and addressed. Perhaps the conclusion section could be shortened, and focused to the most important outcomes of the study, e.g. better performance of singlepol states vs dualpol for some Ice types.

Due to the above given comments, I will suggest your paper for further publication in the Remote Journal.

Specific comments:

Page 3, line 125. I would replace “distance” by “range”.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

The article An Improved Sea Ice Classification Algorithm with Gaofen-3 Dual Polar-ization SAR Data based on Deep Convolutional Neural Networks an important  issue is considered. The authors presented an interesting study. After carefully reading the manuscript, I would suggest making certain changes.

Detailed recommendations are provided below.

1. Introduction–a uniform style of references to literature is needed
2. Introduction - the question of choosing a satellite for research remained unsolved. 
3. (L 107) FSII - a  more detailed mode
4. (L333) Figure 5.- Should be recommended to change view, some parts are too small.
5. It would be interesting for the reader to see a comparative analysis of the methodology on different satellite data.

Author Response

Please see the attachment

Author Response File: Author Response.docx