A Stereo Matching Method for 3D Image Measurement of Long-Distance Sea Surface
Round 1
Reviewer 1 Report
This paper is about stereo reconstruction of sea surface for tsunami prediction. Taj at the first review I found the idea really interesting and promising, but the authors did not make several comments I asked for. Practically most of them were editorial and a few related to the method namely:
The authors did not get the article to the classical layout. There are too many paragraphs and some of them are unnecessary e.g:
Line 87 to 93
line 133-136 the same.
There are too many references to the description in the lines or that it will be described in the line, the text is thus unreadable. Please describe the research method in one paragraph step by step, stage by stage.
There was added according to my suggestion a paragraph about the experiment but the authors did not refer to the question about using the geodesic distance called Path distastce which takes into account the curvature of the earth. According to the authors research in their solution and the solutions compared in tab 2 Euclidean distance method's. Please refer to this in your review response.
A little misunderstood that the reference to fig 13 and 14 are in a different paragraph than the figures . I suggest setting this up chronologically.
Please refer to the study in the context of day and night, season, etc. Did the studies that are being compared take place under the same conditions ? Did the tests take place in good weather ? Does the weather affect the performance of the algorithm or was it tested?
The tsunami problem is very serious so the proposed solution must examine many elements ( e.g.: surface elevations etc.) affecting the system. Please refer to the real time and alarming capabilities of the system.
Author Response
Please see the attachment
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The revised paper has become more clear now and well-structured.
I thank the authors for the attention to all my comments and taking them into account.
I recommend the paper for publication after few corrections concerning the following uncertainties and misprints (numbers are the numbers of lines).
- “Experiments are conducted on sea surface images with different shooting times and distances, which verifies the effectiveness of the presented method” - Experiments themselves do not verify the effectiveness of the method. Their results can.
- measures –> measurements?
- Indian -> India?
Fig.1 The waves here are not shown correctly. According to the dispersion relation, the smaller the water depth, the larger the wave height. The wave closer to the shore should be higher than the previous one.
- Why still water is an obstacle? Synchronization? Exposition? At long distances even 1/24s exposition is enough not to smoothen the image.
- losses -> loses?
- Please add some words about the Discussion section.
Table 1. What is the measure of “the difference between the two matched points”? Pixels, meters, etc?
Do “the two matched points” mean a pair of points (left-right) for one object or two neighbor objects?
- “The measurement of tsunami requires a smaller than 8 pixels matching error” – I still do not understand this (it is mentioned as a general fact). If you have cameras with two times better resolution, is it still 8 pixels or 16?
444-446, Fig. 15. A colorbar would be helpful here.
450 “the intensity where there is a sea wave is larger than the intensity where there is the background.” This phrase is difficult to read. I would suggest “the intensity in the presence of a sea wave is larger than the background intensity”.
Fig. 16. A second subplot with reconstructed sea surface profile (if it is possible at this stage of the study) would be appropriate here to illustrate the adequacy of the method.
487-489. “As the tsunami waves slow down near the coast, the waves behind will superimpose onto the waves in front, resulting in a significant increase in sea surface height near the coast.” It is not really so. Wave height increasing is related not to wave superposition but to the energy accumulation inside smaller area than in the open sea (the wavelength is becoming shorter while the wave energy is the same).
513 “consumption consumption”
480-519. Suggest to combine it to one section “Discussion and Conclusion” as the discussion is still close to conclusion.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Thank you for providing this new version of your paper. The authors have incorporated my suggestions into the new version of the manuscript.
The layout of the document and the large number of sections do not meet the requirements of the journal, and some of them could be removed to achieve the classic style of a modern scientific publication, but this does not reduce the level of work and this aesthetic sense of the reviewer. It is a pity that the translations and answers to the scientific questions described in the answer for the reviewer were not included in the discussion. This would considerably enrich the article with scientific considerations on applicability, the direction of future work, and would allow for a critical look at the proposed solution. I think the authors agree with my opinion that they have prepared clear but convincing answers to the questions that arose after reading the first draft of the article. I regret that it was only in the second version that I received answers to my questions. They convinced me, and therefore I believe that they should be included in the article as a critical scientific discussion. In my opinion, the article will be ready for publication.
Author Response
Please see the attachment
Author Response File: 
Author Response.docx
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
The issue dealt with in this paper is of practical interest. The paper addresses the problem of tsunami detection in real time and in all areas by relying on the measurement of 3D images at distances of 4 to 20 km (and measurements in bad weather). In order to evaluate the performance of the developed methodology, the experiments are carried out on images of the sea surface with different shooting times and distances. The problem, the objectives and the adopted methodology were well presented. However, it would be useful to strengthen the content of the paper taking into account the various comments indicated below and to provide some clarifications to the questions given below.
- The general introduction should be strengthened by further specifying the practical interest of the work carried out and its implementation in an operational context, compared to other alternatives recognized in the literature.
- The methodology developed is based on data and algorithms requiring significant algorithmic complexity. In this context, it would be useful to further specify the algorithmic complexity of the methodology developed and also that of the algorithms used. And to explain well the ways allowing the possible reduction of this complexity. In particular, in the experimentation part, it would be important to specify the calculation time necessary for the execution of the developed methodology.
- It would be useful to specify the limits of the methodology developed, taking into account the meteorological conditions, the geography of the area considered, the sensors used, etc.
- It would be important to specify the implementation of the methodology developed in a practical and operational context.
- The equations must be checked, the quality of the figures is not optimal.
- The conclusion and the perspectives must be reinforced with the presentation of a complete assessment of the work carried out including the limits of the methodology developed, and also the open research themes with elements allowing the mastery of the orientations presented.
Reviewer 2 Report
First of all, I would like to say that this manuscript focuses on a very interesting research problem. The article covers the topics included in the main subjects of Journal of Marine Science and Engineering and I recommend to be considered for the publication after a revision.
TITLE
The article’s title is suitable with the content of the paper.
ABSTRACT
The abstract is well-designed and briefly express the present research thus being of interests and readable thus capturing the reader’s attention. It present in an appropriate manner the main research hypothesis, the problem statement, the methods and the main findings.
KEY WORDS
The key words are appropriate to the present research and are clearly stated.
ORIGINALITY
The article meets a high level of originality argued by the main research theme and the research hypothesis. Furthermore, the originality of the paper is highlighted by the main results of the paper.
The authors construct a well-designed theoretical background closely related to the current specialised literature in the field. A short recommendation I would like to made, it is stated in the final part of this review form.
THE PAPER S STRUCTURE
The structure of the paper is correct in line with the journal standards and meet the publication requirements considering the paper logic. The objectives seem to be clear formulated as well as the investigation is drawn. The core argument of the paper illustrates the paper relevance and the research originality. The results are clearly express and well connected both to the theoretical framework and discussions.
THE METHODS
The methodological design is appropriate and the methods fit well to the present investigation.
ONCLUSIONS
the conclusions fit well summarising the main ideas of the present analysis.
THE GRAPHICAL SUPPORT
The graphical support is well formatted, appropriate illustrating the text content.
THE ENGLISH LANGUAGE
I think the English is ok as far as I could see. I enjoyed to read this paper in English and the language seems well but I think that an opinion of a native English speaker is welcomed. In other words, if the authors used a specialised proofreading services and they could prove this aspect I trust the opinion and the work of this proofread service. On the other hand, I put my trust regarding the English language on the journal editors but I repeat the language seems well.
RECOMMENDATIONS
1. Poor review of the literature and few items containing other proposed solutions or similar but different solutions to the problem proposed by the authors. If the problem is not addressed in journals of international standing, perhaps it is not important enough to be addressed in a scientific journal? Therefore, I think the introductory review is too weak and needs improvement.
2. The text from the abstract and the introduction cannot be the same. It is imperative that this be improved.
3. Why do the authors use only Euclidean distance in the method?
Is it because in the cited method was so or is the reason that they do not take into account the curvature of the Earth?
Do the authors have any experience with using other types of distance measurements?
Finally, I recommend the publication of this paper with some minor revision considering the above mentioned aspects, references and citations.
I want to see the revised version of this paper before publication for a final acceptance and to ensure that the revision has been completely and carefully made.
Reviewer 3 Report
The paper is devoted to stereo reconstruction the sea surface in the far zone (4-20 km) in order to predict tsunami. The idea is really interesting and promising, but the authors do not concern the precision of the final result – what is the method sensitivity to the surface elevations typical for tsunami wave? What is the accuracy of retrieved tsunami wave height and location? What is the criterion for the presence of tsunami wave? It is questionable, because stereo vision at sliding angles can give very large coordinate errors at long distances, and very accurate calibration is needed. I see that tsunami is a quite rare event to test this, but at least the accuracy can be tested on large far objects, like ships, islands, etc.
In this context, the work cannot be positioned as an alternative to the existing tsunami prediction systems until its efficiency is proved. Reading the Abstract, it seems that the authors suggest the ready tsunami prediction system, which is not true. It should be emphasized that only the part of stereo reconstruction procedure is presented (stereo matching), and more details about the system and methods should be given instead of criticism of DARTII, etc, or describing the techniques which are not discussed in the present paper (lines 9-12).
Nevertheless, the stereo matching procedure description is convincing. I suggest only to add the example of 3D surface reconstruction to clarify if the suggested technique is really suitable for the purpose the authors set.
My other concern is the paper organization.
Introduction contains the information that is difficult to follow without reading the whole paper. There are many abbreviations and terms that can be unclear for a common reader (cost volume, descriptor, disparity, penalty, …). It would be helpful if the authors provide a table of symbols and definitions. But in any case, the details starting from line 51, references to the figures and schemes can be given in another section (say, Section Methods). Introduction lacks literature review. Are there any similar stereo systems for the sea surface measurements? E.g., an efficient video observational technology for in field wave measurements was suggested and wildly implemented by Fedele, Benetazzo et al.
A separate Section can be devoted to the experimental set up – locations, geometry, camera parameters, calibration procedure, etc.
There is no conclusion Section, but in fact, Section Discussion is not a “discussion”, but “summary and conclusion”. Discussion is expected to clarify the questions listed above about the method applicability for tsunami prediction.
On the whole, I suppose that the paper is worth publication after improvement concerning the mentioned issues.
Minor comments:
- “in every area” – what areas exactly?
- Space after [2]
- Space before [3]
- “to increase speed, the searching path is decreased to 8” – it is not clear (before reading the following Sections) what means the value “path=8”
- What is RANSAC+SURF method? Provide a reference please
- “The measurement of tsunami requires smaller than 8 pixels matching error” – Does not it depend on camera resolution, viewing area, etc?
93, 94. “stratagy” – strategy
- “The descriptor is a vector combined of sea wave’s appearance features” – it is not a strict and clear definition. What is the measure of features (e.g. their size, coordinates, …)? It is discussed further, but here this definition may confuse a reader.
112-113. “the problem of sea wave matching can be convert to the problem of matching between two feature vector sets L, R”. – It should be mentioned that this is correct only in a case when the two images are taken with similar camera’s geometry. Generally, the feature pixel sizes and orientations can be different in left and right photo, depending on camera positions, angles, zoom, etc.
- “convert” -> converted
- “matching precision is 100%” – what is the criterion of matching precision? It can be understood only few pages later.
- The whole paragraph. “We assume the earth is a sphere” is mentioned twice
- (9), (10) How are the correctly and wrongly detected correspondences defined? Manually (visually) it is possible to define them only with few pixels precision, and XYZ reconstruction can be biased. Thus, 100% precision is not indeed “100%” but can contain some error.
- “litter bit” – “little bit”?
Fig. 12, 13. There is no legend explaining what color each method corresponds (it is present in the text, but can be repeated for convenience)
Fig. 12,13 The numbers and axis notations are too small. Please increase the font size. The numbers in zoomed areas on Fig.13 are undistinguishable at all.
Fig. 13. I don’t see the green line mentioned in the text.
- “Enlarging the local area of these images (see the small gray local enlarged images), we find our
proposed method…” – it is not seen from the figure
- “Teddy on Middlebury Stereo Pages”. Please provide a reference here
Fig. 16. (a-b) are “disparity maps” and (c-f) are “dense maps”. What is the difference?
