Concrete Crack Identification and Image Mosaic Based on Image Processing
Round 1
Reviewer 1 Report
In the manuscript, a procedure for crack identification based on Image Processing is proposed. The object distance method and the scale method are compared for the aim of estimating the crack width. A double edge pixel statistical method is applied to calculate the crack length. Finally, a slightly modified version of an image mosaic algorithm is used for splicing and stitching crack images.
The topic is very interesting and relevant, yet the Paper needs to be importantly improved. Specifically:
- Conversion to greyscale and correction of lens distortion are very basic image preprocessing tools, which can be omitted for brevity sake.
- As above, Canny edge algorithm is widespread and common knowledge; its combination with the Otsu method and any algorithm for the removal of the remaining, unrelated noise effects via estimation of the connected field is not at all a novelty (see, for instance, the mentioned paper Yang J G , Li B Z , Chen H J . Adaptive Edge Detection Method for Image Polluted Using Canny Operator 448 and Otsu Threshold Selection[J]. Advanced Materials Research, 2011, 301-303:797-804, and Fang, M., Yue, G., & Yu, Q. The study on an application of otsu method in canny operator. In Proceedings. The 2009 International Symposium on Information Processing, ISIP 2009, p. 109, Academy Publisher)
- The mentioned “double edge skeleton method” is reported without any related source in bibliography.
- the Scale-invariant feature transform (SIFT) algorithm for image stitching is very widely known and commonly utilised in the image processing community nowadays (as even the Authors explicitly say at page 8, line 226; see, for instance, Cheung, W., & Hamarneh, G. (2007, April). N-sift: N-dimensional scale invariant feature transform for matching medical images. In 2007 4th IEEE International Symposium on Biomedical Imaging: From Nano to Macro (pp. 720-723). IEEE.; Lindeberg, T. (2012). Scale invariant feature transform; and Li, Q., Wang, G., Liu, J., & Chen, S. (2009). Robust scale-invariant feature matching for remote sensing image registration. IEEE Geoscience and Remote Sensing Letters, 6(2), 287-291).
- The feature point correlation comparison and priority matching algorithm, which seems to be main point of this paper, is too briefly discussed as can be seen, e.g., in page 9, lines 252 to 254: “The second step is to fine tune the alignment. Move and rotate the right image so that as many feature points as possible can be overlapped with the same correlation degree. This step achieves maximum splicing.” This explanation lacks some technical detail and does not allow reproducibility. Moreover, the specific splicing process is supposedly shown in Figures 13 to 15; yet they only report the final results without showing or explaining the actual procedure.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The authors presented an approach for crack detection and characterization in concrete. The paper is well organized and the procedure is well described. In addition they consider many aspects to improve the image and the results achieved are quite promising.
My major comments is related to the presentation of the novelty and improvements beyond the state of the art of the procedure presented. It is indeed not clear in the introduction how they overcome the limits complained about the other works made.
Author Response
Dear reviewer:
I am very grateful to your comments for the manuscript. According with your advice, we amended the relevant part in manuscript. Some of your questions were answered below.
Point 1: The authors presented an approach for crack detection and characterization in concrete. The paper is well organized and the procedure is well described. In addition they consider many aspects to improve the image and the results achieved are quite promising.
My major comments is related to the presentation of the novelty and improvements beyond the state of the art of the procedure presented. It is indeed not clear in the introduction how they overcome the limits complained about the other works made.
Response 1: I really appreciate your confirmation of the manuscript. More than anything, your comments have played a vital role in the improvement of the manuscript. To this end, we add this part in the last two paragraphs of the introduction to supplement the background information. In addition, we also improved the details of the experiment and enriched the conclusion. Especially in the last part of Sec. 2. Thank you again for your positive comments and valuable suggestions toimprove the quality of our manuscript.
Author Response File: Author Response.pdf
Reviewer 3 Report
The manuscript regards determination of crack sizes of concretes using computer vision systems.
The structure of the manuscript is incorrect. The content is mixed – theories and literature information are mixed with authors’ findings. Therefore, the content of the sections should be ordered and some paragraph moved to other places. The recommended structure is described at https://www.mdpi.com/journal/applsci/instructions .
It is very difficult to guess what is the authors’ contribution and what is taken from the literature. I recommend using phrases like: “we proposed...”, “the method is adopted after...”. I cannot guess whether e.g., Fig. 9 is your own contribution or is a typical approach.
The manuscript looks more like a textbook than as an article. You should not write everything that is connected with the topic. You should rather give a short explanation and refer to literature. Therefore, you should compress your theoretical part.
The manuscript lacks good scientific style. There are many too long sentences. I recommend read and apply some of the guides regarding scientific writing, such as that: https://cgi.duke.edu/web/sciwriting/index.php or https://www.elsevier.com/connect/writing-a-science-paper-some-dos-and-donts Section “Avoid long sentences”: ``Nowadays, the average length of sentences in scientific writing is about 12 to 17 words’’ – some sentences have over 40 words!
The number of significant figures expressing errors is incorrect. The modern metrological rule says that uncertainties are reported with a maximum of two significant figures. I cannot find any justification for giving 4 figures describing errors. Additionally, when the numbers are correctly formatted, the reader can better understand the main idea.
According to the SI ( https://www.bipm.org/utils/common/pdf/si-brochure/SI-Brochure-9-EN.pdf ) symbols for quantities should be printed in italic. So, you have to change all the variables used in the manuscript.
In Sec. 2 you presented how works the algorithm. Please extend it showing how it works at the ends of the cracks, similar to Figs. 2–6. Additionally, show what the algorithm does with small cracks.
Sec. 3: Your approach for determination of length is discussable. It is similar to the problem of determining the length of a sea-coast – the accurate measurement, the longer is the coast. You should refer to an accepted standard, such as ISO.
The style of referencing to equation put after the text is incorrect. Instead of e.g., “is defined as equation (16):” you should simply write “is defined as:”. Of cause, you should change all similar cases.
Your section entitled “Discussion” is rather “Conclusion”. Consider building a better structure of the document.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
Your article is really good. But the results in the laboratory are always of higher quality than in situ results. The effect of technical seismicity in situ reduces the quality of the results and cannot be practically eliminated. The error increases as the distance between the camera and the structure increases.
Author Response
Dear reviewer:
I am very grateful to your comments for the manuscript. According with your advice, we amended the relevant part in manuscript. Some of your questions were answered below.
Point 1: Your article is really good. But the results in the laboratory are always of higher quality than in situ results. The effect of technical seismicity in situ reduces the quality of the results and cannot be practically eliminated. The error increases as the distance between the camera and the structure increases.
Response 1: Thank you for your positive comments and valuable suggestions of our manuscript. More importantly, your comments are exactly what we are thinking about, and they are also of great concern to current experimenters. Indeed, due to the ideal environmental conditions in the laboratory, the calculation accuracy is definitely higher than in the field. We also made some considerations regarding the influence of field test conditions. For example, the use of unmanned aerial vehicles (UAV) can ensure the absolute hover state during flight to take photos, which can solve the vibration problem and improve the image quality. A camera with an adjustable focal length obtains a higher quality image by adjusting the focal length when the working distance is longer. But there are still many unknown and harsh conditions that need to be overcome. In view of these unknown adverse conditions, we are perfecting the experimental scheme to be put into field test as soon as possible.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
see attachment per review report
Comments for author File: Comments.pdf
Author Response
Dear reviewer:
I am very grateful to your comments for the manuscript. According with your advice, we amended the relevant part in manuscript. Some of your questions were answered below.
Point 1: Conversion to greyscale and correction of lens distortion are very basic image preprocessing tools, which can be omitted for brevity sake.
Response 1: We feel great thanks for your professional review work on our article. According to your nice suggestions, we have made extensive corrections to our previous manuscript and deleted this part.
Point 2: As above, Canny edge algorithm is widespread and common knowledge; its combination with the Otsu method and any algorithm for the removal of the remaining, unrelated noise effects via estimation of the connected field is not at all a novelty (see, for instance, the mentioned paper Yang J G , Li B Z , Chen H J . Adaptive Edge Detection Method for Image Polluted Using Canny Operator 448 and Otsu Threshold Selection[J]. Advanced Materials Research, 2011, 301-303:797-804, and Fang, M., Yue, G., & Yu, Q. The study on an application of otsu method in canny operator. In Proceedings. The 2009 International Symposium on Information Processing, ISIP 2009, p. 109, Academy Publisher)
Response 2: Thank you for your suggestion. Indeed, this method is not our innovative content. Our innovation lies in the comparison of the two calculation methods, using the double edge skeleton method to calculate the crack length and the improvement of the crack stitching algorithm. Therefore, we have compressed this section according to your suggestion. Of course, I have read the document you mentioned carefully and put it in the manuscript.
Point 3: The mentioned “double edge skeleton method” is reported without any related source in bibliography.
Response 3: I am very sorry, we have not expressed this clearly in the manuscript. We use this method to calculate the crack length and call it “double edge skeleton method”, which addresses the drawbacks of the general “central skeleton method”. Thank you very much for your kind reminder, which we have added in the last paragraph of the introduction.
Point 4: The Scale-invariant feature transform (SIFT) algorithm for image stitching is very widely known and commonly utilised in the image processing community nowadays (as even the Authors explicitly say at page 8, line 226; see, for instance, Cheung, W., & Hamarneh, G. (2007, April). N-sift: N-dimensional scale invariant feature transform for matching medical images. In 2007 4th IEEE International Symposium on Biomedical Imaging: From Nano to Macro (pp. 720-723). IEEE.; Lindeberg, T. (2012). Scale invariant feature transform; and Li, Q., Wang, G., Liu, J., & Chen, S. (2009). Robust scale-invariant feature matching for remote sensing image registration. IEEE Geoscience and Remote Sensing Letters, 6(2), 287-291).
Response 4: The Scale-invariant feature transform (SIFT) algorithm is indeed a very good stitching algorithm. We are just the application and improvement of this algorithm. The papers you mentioned all gave a good introduction to the algorithm, which we read carefully and put in the manuscript.
Point 5: The feature point correlation comparison and priority matching algorithm, which seems to be main point of this paper, is too briefly discussed as can be seen, e.g., in page 9, lines 252 to 254: “The second step is to fine tune the alignment. Move and rotate the right image so that as many feature points as possible can be overlapped with the same correlation degree. This step achieves maximum splicing.” This explanation lacks some technical detail and does not allow reproducibility. Moreover, the specific splicing process is supposedly shown in Figures 13 to 15; yet they only report the final results without showing or explaining the actual procedure.
Response 5: The second step of fine tune the alignment is achieved by adjusting the number of matching points. Details of this part have been added to the manuscript according to your nice suggestions. And we have also added the detailed stitching process to the manuscript about Figures 13 to 15. we were really sorry for our careless mistakes. Thank you for your reminding.
Author Response File: Author Response.pdf
Reviewer 3 Report
In the revision, authors have taken into consideration most of my remarks. The quality of the manuscript is much better now. I recommend the paper for publication as it is.
Author Response
Dear reviewer:
I really appreciate your confirmation of the manuscript. More than anything, your comments have played a vital role in the improvement of the manuscript.
Thank you again for your comments and valuable suggestions toimprove the quality of our manuscript.
