The Surface Crack Extraction Method Based on Machine Learning of Image and Quantitative Feature Information Acquisition Method

Round 1

Reviewer 1 Report

Dear authors,

Here are my comments:

Fig. 1. – at least the disputable areas like Taiwan and islands in the South Chinese Sea have to be special marked or omit. The scientific paper is not place for favour side in any political conflict. Scales for a, b and c.

 

Fig. 3. – scales are missing. To understand what is scale of aerial exploration they need to be on figure(s). What is difference between bare ground, green vegetation and soil crust? The variable for distinguishing must be clearly defined, not qualitative.

 

Leave one out method is not eplained correctly. One data is left and its value is estimated from the rest. Those are not „k“ new datasets, but variation of original one. Permutation and bootstrap are not same procedures. Each of them must be better presented on real data.

 

This statemaent is not clear: „The idea of this algorithm is to extract the peripheral contour of the target, and then use the contour to corrode the boundary of the target image until it can no longer be corroded. The crack skeleton was extracted by K3M method, and the result is shown in Figure 4.“. Explain why you applied skeleton algorithm on your data, and what are advantages.

 

„The purpose is to break the branching and intersecting cracks, so that there are only two node types of endpoint and junction-point in the skeleton pixels, as shown in Figure 7.“ – please explain why this is neccessary to do for interpretation of real ground cracks. Is it computing condition for crack length calculation?

 

Crack width calculation – how the process is reliable after application of skeleton and burr processes.

 

Table 2. – how did you obtain correct classifications? Fieldwork?

 

Fig. 16 – scales? Why you did not emphasise (mark) cracks on (a)?

 

Fig. 17 – the marked cracks are not the most visible and prominent in the middle of the map on Fig. 16. Why?

 

Conclusion – support with real numbers from your study like success rate, number of data, size of analysed areas, type of soil…

 

Plagscan results (similarity check) shows:

13.1 % similarity with paper "A New Identification Method for Surface Cracks from UAV Images Based on Machine Learning in Coal Mining Areas" (doi:10.3390/rs12101571).

 

Kind regards,

reviewer

 

Author Response

Response to Reviewer 1:

Thank you so much for your comments regarding our manuscript entitled " The surface crack extraction method based on machine learning of image and quantitative feature information acquisition method " by Fan Zhang et al. submitted to Remote Sensing. We have revised the manuscript according to the reviewer’s comments. The detailed response is in the attached document.

Author Response File: Author Response.docx

Reviewer 2 Report

This manuscript proposed a novel method to quantify the surface crack based machine learning and quantitative feature acquisition method. The proposed method can effectively extract the crack features including length, width, direction, location, fractal dimension and number, and crack and dispersion rates. Finally, the images taken by the UAV were used to validate the effectiveness of the proposed method with promising results. Overall, the topic of this research is interesting. However, the manuscript was not well organized and written. It requires a further modification before accepted for publication in Remote Sensing. The detailed comments are given as follows.

1. The title is not suitable. From my point of view, the proposed method is general method that can be applied in different types of images not only UAV images.
2. There are a large number of similar research in this area. Please illustrate the main innovation of this study. Why was machine learning-based method selected for the task of interest? How about conventional image processing methods?
3. Line 72, please give where the images were taken. Throughout the manuscript, the authors just mentioned they are surface cracks. However, further details are required. E.g. building surface, bridge surface or pavement?
4. Figure 2: please change the font in the figure to satisfy the journal requirement.
5. The authors used SVM as machine learning method. Please explain why SVM was considered. The deep learning method seems more advantageous.
6. Please add the brief introduction on the fundamental of SVM. I suggest that authors can read the following references and include it in the manuscript:

 https://doi.org/10.3390/rs13020240

https://doi.org/10.1016/j.conbuildmat.2018.08.011

https://doi.org/10.1016/j.dib.2018.11.015

7. How did the authors set the hyperparameters of SVM? Different settings of parametres can lead to remarkably different performances of the trained model.
8. The authors should add the training and validation results to prove the effectiveness of the trained SVM model.
9. The authors mentioned that leave-one-out method was used during the model training. However, from the description, it looks like k-fold cross-validation. Please have a check.
10. Figure 15(b): different types of lines (straight, dotted, etc) are suggested to distinguish different results.
11. More future research should be included in Conclusion part

Author Response

Response to Reviewer 2:

Thank you so much for your comments regarding our manuscript entitled " The surface crack extraction method based on machine learning of image and quantitative feature information acquisition method " by Fan Zhang et al. submitted to Remote Sensing. We have revised the manuscript according to the reviewer’s comments. The detailed response is in the attached document.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Dear authors,

You addressed all my comments, only scales not. As I consider them pretty important for readers it is why I evaluated this version as - minor review. Please, add them.

Regards,

reviewer

Author Response

Thank you so much for your comments regarding our manuscript entitled " The surface crack extraction method based on machine learning of image and quantitative feature information acquisition method " by Fan Zhang et al. submitted to Remote Sensing. We have revised the manuscript according to the reviewer’s comments. 

Author Response File: Author Response.docx

Reviewer 2 Report

The comments have been addressed and the paper can be accepted in present form.

Author Response

Thank you so much for your accept regarding our manuscript entitled " The surface crack extraction method based on machine learning of image and quantitative feature information acquisition method " by Fan Zhang et al. submitted to Remote Sensing.