Improvement and In-Situ Application of an Evaluation Method of Ballasted-Track Condition Using Digital 2-D Image Analysis
Round 1
Reviewer 1 Report
The theme of this study is timely and interesting to the readers. However, it seems necessary to describe the preconditions for drawing conclusions in more detail for the publication. For example, the following points can improved.
1) 2.2 Improvement of image analysis method
It is concluded that changing the background color and changing the algorithm led to improvement, but it is unclear which contribution is greater. This is useful information for the readers, so please add an explanation.
2) 3.2.1 Applicability of shape indices
Abrasion is evaluated using the relationship between the shape index and abrasion referring a previous study, but this relationship is important, so please explain it in detail in this paper.
3) Figure 6, Figure 7
Although linear approximation is performed in each figure, non-linear approximation seems to be appropriate instead of linear approximation. Please explain the premise that linear approximation is possible.
Author Response
1) 2.2 Improvement of image analysis method
It is concluded that changing the background color and changing the algorithm led to improvement, but it is unclear which contribution is greater. This is useful information for the readers, so please add an explanation.
⇨ The most important step in image analysis method is to separate ballasts and background. This study tried to develop the best method to separate ballast aggregates from background. To do so, the authors found out that specific algorithm with chromatic color background is the simplest and the best way. The algorithm is to convert RGB image into HSV color model and use S-channel (Saturation) of HSV color model. HSV color model is frequently used in computer vision. This method enhances the separation possibility to 100 %, comparing with the previous method of 50 to 100%.
This Brief explanation is added in chapter 2.2.
2) 3.2.1 Applicability of shape indices
Abrasion is evaluated using the relationship between the shape index and abrasion referring a previous study, but this relationship is important, so please explain it in detail in this paper.
⇨ Relationships between shape indices and abrasion loss from previous study are added in chapter 2.1.
3) Figure 6, Figure 7
Although linear approximation is performed in each figure, non-linear approximation seems to be appropriate instead of linear approximation. Please explain the premise that linear approximation is possible.
⇨ Abrasion loss versus TQI or number of tamping is proportional relation. If abrasion loss is low, TQI or number of tamping is also low and vice versa. However, coefficient of determination is relatively small because the number of sampling points is small. In case of non-linear approximation like exponential relation, the results are not much different in linear. As a further study, this limitation could be complemented with more field sampling and other approximation method.
Author Response File: 
Author Response.docx
Reviewer 2 Report
General
The paper is a good example of relevant applied science. I have only a few comments.
It would be of interest for the reader to know the type of rock. Not a petrographic description just a sentence or two. If there is a requirement for the L.A. value in Korea that could be mentioned in the introduction.
Abrasion loss is so central to the content of the paper that is has to be described. It is not sufficient to refer to other sources.
Specific comments
Lines 32 to 42. The requirement of 20% passing the 22.4 mm sieve is mentioned twice.
Line 60. I suggest to take away the text from: “the gravels used for railways are larger than sand or soil particles, thus”.
Line 70. Move the reference [10] from the heading to the text.
Line 155. Should It be “… can be printed or presented on a screen.”?
Line 182. I would suggest something like: “….abrasion loss based on the L.A. test versus shape indices, TQI…”
Line 187. “….shows if the overall shape…”
Line 202. “may have been” is to vague. It is better to state the correlation is low.
Line 219. Should be “..comparing with…”.
Line 253. Is it a rate? I think it could be “table show that 20% fine passing occur between 500 and 1000.”
Lines 255 to 258 I find these lines difficult to follow at the same time they are very crucial for the main conclusions of the paper. It is important that this is formulated in a way that it is easy to follow and persuade the reader that this can be something to apply in practice.
Line 258. 10.57% is not an approximation I believe it is more exact than what the method can give. I would suggest 10% as an approximation that can be used as guide line.
Line 273. Change to “…exhibited a low correlation…”
Author Response
The paper is a good example of relevant applied science. I have only a few comments.
It would be of interest for the reader to know the type of rock. Not a petrographic description just a sentence or two. If there is a requirement for the L.A. value in Korea that could be mentioned in the introduction.
⇨ A kind of ballast aggregates in Korea is rhyolitic tuff. In korea railroad corporation standard, there are requirements for ballast aggregates properties. Ballast size is between 22.4 ~ 63 mm and abrasion loss is under 25%. These are added in introduction.
Abrasion loss is so central to the content of the paper that is has to be described. It is not sufficient to refer to other sources.
⇨ Abrasion loss calculation after L.A. abrasion test is added in chapter 2.1.
Specific comments
⇨ Everything was fixed.
Author Response File: Author Response.docx
Reviewer 3 Report
- A large number of linear regression curves are presented but they do have very weak correlation with the data as evidenced through the low values of R2. I suggest a nonlinear regression analysis can help to provide better empirical relationships.
- The data presented in "Figure 11. Sieve analysis results" is not clear due to lack of information on the different surface's and mid's used in the legend.
- Authors discuss the development of image analysis software. Provide more details. Which software or coding platform was used and whether this comes under the technical contributions of this study. I think this section 2.2.4 with addition of more details can be shifted to a separate appendix.
- Consider improving the research background. Some of the following references can be useful to include the recent state of the art knowledge pertaining to this field.
Le Pen, L. M., Powrie, W., Zervos, A., Ahmed, S., & Aingaran, S. (2013). Dependence of shape on particle size for a crushed rock railway ballast. Granular Matter, 15(6), 849-861.
Sun, Y., Indraratna, B., & Nimbalkar, S. (2014). Three-dimensional characterisation of particle size and shape for ballast. Géotechnique Letters 4(3), 197-202.
Ouhbi, N., Voivret, C., Perrin, G., & Roux, J. N. (2016). Railway ballast: grain shape characterization to study its influence on the mechanical behaviour. Procedia engineering, 143, 1120-1127.
Author Response
- A large number of linear regression curves are presented but they do have very weak correlation with the data as evidenced through the low values of R2. I suggest a nonlinear regression analysis can help to provide better empirical relationships.
⇨ Abrasion loss versus TQI or number of tamping is proportional relation. If abrasion loss is low, TQI or number of tamping is also low and vice versa. However, coefficient of determination is relatively small because the number of sampling points is small. In case of non-linear approximation like exponential relation, the results are not much different in linear. As a further study, this limitation could be complemented with more field sampling and other approximation method.
- The data presented in "Figure 11. Sieve analysis results" is not clear due to lack of information on the different surface's and mid's used in the legend.
⇨ There are too many ballast sampling points so it is hard to distinguish each other in one graph. Sieve analysis is not a main research subject so this study focused sieve analysis results according to sampling depth (surface versus mid depth).
- Authors discuss the development of image analysis software. Provide more details. Which software or coding platform was used and whether this comes under the technical contributions of this study. I think this section 2.2.4 with addition of more details can be shifted to a separate appendix.
⇨ More details about analysis software(programming language, used library) are added in chapter 2.2.4.
- Consider improving the research background. Some of the following references can be useful to include the recent state of the art knowledge pertaining to this field.
Le Pen, L. M., Powrie, W., Zervos, A., Ahmed, S., & Aingaran, S. (2013). Dependence of shape on particle size for a crushed rock railway ballast. Granular Matter, 15(6), 849-861.
Sun, Y., Indraratna, B., & Nimbalkar, S. (2014). Three-dimensional characterisation of particle size and shape for ballast. Géotechnique Letters 4(3), 197-202.
Ouhbi, N., Voivret, C., Perrin, G., & Roux, J. N. (2016). Railway ballast: grain shape characterization to study its influence on the mechanical behaviour. Procedia engineering, 143, 1120-1127.
⇨ added
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The reviewer see that he paper can be accepted as in the present form.
Author Response
Thank you for reviewing my paper.
Reviewer 3 Report
I have checked the authors responses but the associated revisions in the manuscript are not apparent, including the addition/revision of text as well as addition of references to improve the background information. Please send us a list of all the changes you have made, citing manuscript pages, and highlight (or use colour) the changed portion of the manuscript. I am therefore recommending a minor revision to provide authors an opportunity to fix these issues.
Author Response
I uploaded a revised manuscript(.pdf format)
or You can see the changes in my previous revised manuscript(.docx format) with MS Word track change function.
Author Response File: 
Author Response.pdf
