Review Reports - Development of Image Processing for Crack Detection on Concrete Structures through Terrestrial Laser Scanning Associated with the Octree Structure

Round 1

Reviewer 1 Report

There are 6 main issues that the authors must address before this paper can be considered further

The authors need to stream line the paper to better describe the fact that they are using both images (not just imagery processing techniques) and lidar. This needs to be clarified in the title, the abstract and elsewhere. The choice to use the dual data set must be justified as it involves a lot of extra work that is not fully justified in the text.

The technique is only tried on 1 case at 3 different data densities. Changing the data densities should be part of a sensitivity study as part of a Discussion section where the advantages and disadvantages are clearly compared. Move this work to after the results and make it a proper section

Two more cases of different walls must be included in the results section (only present the ones with the best results (see comments above as to changing the densities)

A quantitative means needs to be developed to discuss success and failure

Authors must at least discuss width versus length documentation

The paper lacks a proper literature review. There are very few papers of relevance cited and most are extremely old. Please remove all references that are more than 10 years old unless your project is using a technique that is described in the older paper. Some references of high relevance are listed below.

Relevant algorithms on concrete cracks and terrestrial laser scanning past 10 years

Koch, C., Georgieva, K., Kasireddy, V., Akinci, B. and Fieguth, P., 2015. A review on computer vision based defect detection and condition assessment of concrete and asphalt civil infrastructure. Advanced Engineering Informatics, 29(2), pp.196-210.

Laefer, D.F., Truong-Hong, L., Carr, H. and Singh, M., 2014. Crack detection limits in unit based masonry with terrestrial laser scanning. Ndt & E International, 62, pp.66-76.

Li, T., Almond, D.P. and Rees, D.A.S., 2011. Crack imaging by scanning pulsed laser spot thermography. NDT & E International, 44(2), pp.216-225.

Giri, P. and Kharkovsky, S., 2016. Detection of surface crack in concrete using measurement technique with laser displacement sensor. IEEE Transactions on Instrumentation and Measurement, 65(8), pp.1951-1953.

Kim, M.K., Sohn, H. and Chang, C.C., 2014. Localization and quantification of concrete spalling defects using terrestrial laser scanning. Journal of Computing in Civil Engineering, 29(6), p.04014086.

Valença, J., Puente, I., Júlio, E., González-Jorge, H. and Arias-Sánchez, P., 2017. Assessment of cracks on concrete bridges using image processing supported by laser scanning survey. Construction and Building Materials, 146, pp.668-678.

Anil, E.B., Akinci, B., Garrett, J.H. and Kurc, O., 2013. Characterization of laser scanners for detecting cracks for post-earthquake damage inspection. In International Symposium on Automation and Robotics in Construction and Mining (ISARC), Montreal, QC.

Laefer, D.F., Gannon, J. and Deely, E., 2010. Reliability of crack detection methods for baseline condition assessments. Journal of Infrastructure Systems, 16(2), pp.129-137.

Xu, X., Yang, H. and Neumann, I., 2015. Concrete crack measurement and analysis based on terrestrial laser scanning technology. Sens. Trans. J., 186(3), pp.168-17

Song, M., Yousefianmoghadam, S., Mohammadi, M.E., Moaveni, B., Stavridis, A. and Wood, R.L., 2018. An application of finite element model updating for damage assessment of a two-story reinforced concrete building and comparison with lidar. Structural Health Monitoring, 17(5), pp.1129-1150.

Lõhmus, H., Ellmann, A., Märdla, S. and Idnurm, S., 2018. Terrestrial laser scanning for the monitoring of bridge load tests–two case studies. Survey Review, 50(360), pp.270-284.

Baeza, F.J., Ivorra, S., Bru, D. and Varona, F.B., 2018. Structural health monitoring systems for smart heritage and infrastructures in Spain. In Mechatronics for Cultural Heritage and Civil Engineering (pp. 271-294). Springer, Cham.

Use of octree to process lidar

Wang, J., Lindenbergh, R. and Menenti, M., 2017. SigVox–A 3D feature matching algorithm for automatic street object recognition in mobile laser scanning point clouds. ISPRS Journal of Photogrammetry and Remote Sensing, 128, pp.111-129.

Imagery vs lidar

Olsen, M.J., Kuester, F., Chang, B.J. and Hutchinson, T.C., 2009. Terrestrial laser scanning-based structural damage assessment. Journal of Computing in Civil Engineering, 24(3), pp.264-272.

Additionally, do not use the first person in technical writing

Table 2 usually the quality setting is also reported for the Leica units

For Table 4 since factors D and E do not change they do not need to be in the table and can just be mentioned in the text

Case1 (etc) should appear as Case 1

Table 1 should have an extra column specifying what dataset is being processed

Discussion of TLS is not really necessary; reduce to 1-2 sentences and provide a reference. Also the image is highly idealized so I think it provides the wrong impression and should be removed as it assumes perfect perpendicularity

Line 207 you must say RS data and not just RS (fix elsewhere

Fig. 5 is hard to understand. In the caption be explicit as to the input data.

I could not find a rigorous discussion of the input data. As cracks are of different size, using the number of pixels as a descriptor is not helpful. There should be a size as a function of either crack length or even better crack width

Fig. 8 reorganize so that the entire figure remains on 1 page.

Fig. 11 (and elsewhere) have keys that are unreadable

Fig. 12 why don't you create the octree model first? Please discuss this in the text

Authors must clearly state the limits of the detectable crack lengths and widths and this must be put in context with industrial expectations (eg see AASHTO bridge inspection)

In many places the reference comes too late (eg. octree, k-means, Otsu). All should have a reference where they first appear in the text

I find the use of the term stereoscopic with TLS very weird and would strongly suggest its removal

Author Response

We sincerely thank you for the constructive comments. By virtue of the reviewers’ comments and suggestions, the manuscript has been significantly strengthened both in contents and clarity

Reviewer 1

There are 6 main issues that the authors must address before this paper can be considered further.

1. The authors need to stream line the paper to better describe the fact that they are using both images (not just imagery processing techniques) and lidar. This needs to be clarified in the title, the abstract and elsewhere. The choice to use the dual data set must be justified as it involves a lot of extra work that is not fully justified in the text.

This study did not use both image and lidar data, but image processing on the image extracted from the octree structure of lidar data. We made this clear in the title and abstract..

2. The technique is only tried on 1 case at 3 different data densities. Changing the data densities should be part of a sensitivity study as part of a Discussion section where the advantages and disadvantages are clearly compared. Move this work to after the results and make it a proper section. Two more cases of different walls must be included in the results section (only present the ones with the best results (see comments above as to changing the densities). A quantitative means needs to be developed to discuss success and failure

A quantitative evaluation and comparative analysis of imaging and crack results for three case according to the compression ratio have been performed in terms of True Positive (TP), False Negative (FN), and number of False Positive (FP) in Figure 14 and Table 5. Also, the discussion for characteristics of three cases including the advantages and disadvantages has been included in the manuscript.

3. Authors must at least discuss width versus length documentation

The scope of this paper is mostly on the “detection” of cracks using TLS and has not extended to “quantification” of the detected cracks for widths and lengths. The authors believe the quantification can be performed by employing predeveloped algorithms described in these references.

[References]

55. Liu, C., Tang, C. S., Shi, B., & Suo, W. B. (2013). Automatic quantification of crack patterns by image processing. Computers & Geosciences, 57, 77-80.

56. Adhikari, R. S., Moselhi, O., & Bagchi, A. (2014). Image-based retrieval of concrete crack properties for bridge inspection. Automation in construction, 39, 180-194.

57. Liu, Y., Cho, S., Spencer Jr, B. F., & Fan, J. (2014). Automated assessment of cracks on concrete surfaces using adaptive digital image processing. Smart Structures and Systems, 14(4), 719-741.

58. Liu, Y. F., Cho, S., Spencer Jr, B. F., & Fan, J. S. (2014). Concrete crack assessment using digital image processing and 3D scene reconstruction. Journal of Computing in Civil Engineering, 30(1), 04014124.

To clarify this, a sentence is added in Section 4 with the references as

After detecting the cracks using the proposed method, quantification of cracks can be done by employing existing quantification algorithms mostly based on combination of morphological operations, such as median filtering, morphological opening and closing, skeletonization, edge detection, etc. [55-58].

4. The paper lacks a proper literature review. There are very few papers of relevance cited and most are extremely old. Please remove all references that are more than 10 years old unless your project is using a technique that is described in the older paper. Some references of high relevance are listed below.

Relevant algorithms on concrete cracks and terrestrial laser scanning past 10 years

Koch, C., Georgieva, K., Kasireddy, V., Akinci, B. and Fieguth, P., 2015. A review on computer vision based defect detection and condition assessment of concrete and asphalt civil infrastructure. Advanced Engineering Informatics, 29(2), pp.196-210.

Laefer, D.F., Truong-Hong, L., Carr, H. and Singh, M., 2014. Crack detection limits in unit based masonry with terrestrial laser scanning. Ndt & E International, 62, pp.66-76.

Li, T., Almond, D.P. and Rees, D.A.S., 2011. Crack imaging by scanning pulsed laser spot thermography. NDT & E International, 44(2), pp.216-225.

Giri, P. and Kharkovsky, S., 2016. Detection of surface crack in concrete using measurement technique with laser displacement sensor. IEEE Transactions on Instrumentation and Measurement, 65(8), pp.1951-1953.

Kim, M.K., Sohn, H. and Chang, C.C., 2014. Localization and quantification of concrete spalling defects using terrestrial laser scanning. Journal of Computing in Civil Engineering, 29(6), p.04014086.

Valença, J., Puente, I., Júlio, E., González-Jorge, H. and Arias-Sánchez, P., 2017. Assessment of cracks on concrete bridges using image processing supported by laser scanning survey. Construction and Building Materials, 146, pp.668-678.

Anil, E.B., Akinci, B., Garrett, J.H. and Kurc, O., 2013. Characterization of laser scanners for detecting cracks for post-earthquake damage inspection. In International Symposium on Automation and Robotics in Construction and Mining (ISARC), Montreal, QC.

Laefer, D.F., Gannon, J. and Deely, E., 2010. Reliability of crack detection methods for baseline condition assessments. Journal of Infrastructure Systems, 16(2), pp.129-137.

Xu, X., Yang, H. and Neumann, I., 2015. Concrete crack measurement and analysis based on terrestrial laser scanning technology. Sens. Trans. J., 186(3), pp.168-17

Lõhmus, H., Ellmann, A., Märdla, S. and Idnurm, S., 2018. Terrestrial laser scanning for the monitoring of bridge load tests–two case studies. Survey Review, 50(360), pp.270-284.

Baeza, F.J., Ivorra, S., Bru, D. and Varona, F.B., 2018. Structural health monitoring systems for smart heritage and infrastructures in Spain. In Mechatronics for Cultural Heritage and Civil Engineering (pp. 271-294). Springer, Cham.

Use of octree to process lidar

Wang, J., Lindenbergh, R. and Menenti, M., 2017. SigVox–A 3D feature matching algorithm for automatic street object recognition in mobile laser scanning point clouds. ISPRS Journal of Photogrammetry and Remote Sensing, 128, pp.111-129.

Olsen, M.J., Kuester, F., Chang, B.J. and Hutchinson, T.C., 2009. Terrestrial laser scanning-based structural damage assessment. Journal of Computing in Civil Engineering, 24(3), pp.264-272.

As the reviewer suggests, the introduction has been significantly improved, including all the recommended references for the relevant algorithms on concrete cracks and terrestrial laser scanning, use of octree to process lidar, and imagery vs lidar.

5. Additionally, do not use the first person in technical writing

As the reviewer suggests, the first person expression in all the sentences has been removed.

6. Table 2 usually the quality setting is also reported for the Leica units

The quality setting values of Leica units has been revised in Table 3 and 4.

7. For Table 4 since factors D and E do not change they do not need to be in the table and can just be mentioned in the text

Though the factors A and E do not change for the three Cases (i.e., Original, Case 1, and Case 2), they are also the optimized parameters to catch the cracks at the best. Thus, the factors A and E are kept in Table 4. The use of Otsu’s method is described in section 3.5, so the factor D is removed out from the table as the reviewer recommended.

8. Case1 (etc) should appear as Case 1

They are corrected in Table 4.

9. Table 1 should have an extra column specifying what dataset is being processed

In this paper, the point cloud data obtained by the TLS is visualized as an image form, and it is processed using image processing techniques for crack detection throughout the paper. Thus, the extra column to describe used dataset may not be required. To clarify this, additional sentences are attached in the beginning of Chapter 3 (Line 149) as

The point cloud data obtained by the TLS is compressed by the octree data structure, and it can be converted to an image as shown in Figure 4(a). To detect cracks from the converted image, a series of image processing techniques are used.

10. Discussion of TLS is not really necessary; reduce to 1-2 sentences and provide a reference. Also the image is highly idealized so I think it provides the wrong impression and should be removed as it assumes perfect perpendicularity

As the reviewer suggests, the discussion of TSL has been considerably reduced and the proper references have been added. Also, Figure 2 which seems to have perfect perpendicularity has been removed.

11. Line 207 you must say RS data and not just RS (fix elsewhere)

In Chapter 3.4, RS and CS are changed to RS data and CS data, as the reviewer recommended.

12. Fig. 5 is hard to understand. In the caption be explicit as to the input data.

The additional description of Fig. 5 is described in the manuscript as below.

Figure 5 illustrates how the subtraction finds cracks against the other objects including shading. Illuminance of the original gray-level visualizes objects in the image, and the median filtering removes out sharp illuminance changes that represent small noises or cracks and remains smoothed background image representing concrete surface, large objects, and shading. Then, the subtraction of the original image from the filtered image removes out the background image while remaining objects with sharp illuminance. Since the crack is visualized by very low illuminance profile with shape change, the crack object can be highlighted from the subtracted image, if combined with subsequent image processing to remove out small noises.

13. I could not find a rigorous discussion of the input data. As cracks are of different size, using the number of pixels as a descriptor is not helpful. There should be a size as a function of either crack length or even better crack width.

This study aims at crack identification by comparison with crack map by visual inspection. For quantitative evaluation of accuracy, TP, FN, and the number of TN of the imaging process results were analyzed. We will proceed a rigorous discussion of the input data by comparing with crack map by visual inspection in the future study.

14. Fig. 8 reorganize so that the entire figure remains on 1 page.

Figure 8 has been edited to fit on 1 page

15. Fig. 11 (and elsewhere) have keys that are unreadable

The keys in Figure 11 has been edited for the good readability,

16. Fig. 12 why don't you create the octree model first? Please discuss this in the text

We added the reason not to create the octree model first in the section 2.2

In this study, the 3D Object model (Voxel model) is first generated to easily remove empty nodes by checking empty cells in the sub and left nodes. Thus, it is efficient to implement segmentation and octree in the spatial properties of 3D objects rather than in irregularly distributed point clouds.

17. Authors must clearly state the limits of the detectable crack lengths and widths and this must be put in context with industrial expectations (eg see AASHTO bridge inspection)

As shown in Figure 15(a), the crack widths obtained by the visual inspection are within 0.2mm. According to the Korean structural inspection guide, the crack width less than 0.1mm leads to the grade of the inspected structure as “a”, and width between 0.1mm and 0.2mm leads to “b”. This means the proposed method can be successfully used as a prospective alternative to the conventional visual inspection. This is stated as the end of Chapter 4 (Line 343) as

Though the cracks are not quantified in this study, the crack widths obtained by visual inspection in Figure 15(a) are within 0.2 mm. According to the Korean structural inspection guide [59], the crack width less than 0.1 mm leads to the grade of the inspected structure as “a”, and width between 0.1 mm and 0.2 mm leads to “b”. Thus, the proposed method can be successfully used as a prospective alternative to the conventional visual of large structure with minimal labor and logistic time.

[59] Korea Ministry of Land, Infrastructure, and Transport (MLIT) and Korea. Guideline for Regular and Detailed Inspection, Korea Infrastructure Safety and Technology Corporation (KISTeC); Jinju, Korea, 2017; pp. 6-23.

18. In many places the reference comes too late (eg. octree, k-means, Otsu). All should have a reference where they first appear in the text

The references of what come up first such as octree, k-means clustering, and Otsu has been edited in order.

19. I find the use of the term stereoscopic with TLS very weird and would strongly suggest its removal

The term "stereoscopic" has been removed and the universal description of TSL has been replaced.

Author Response File: Author Response.pdf

Reviewer 2 Report

This is a paper on using TLS data for crack detection on concrete surfaces. Although it provides some interesting results, there is an absolute need to improve the quality of the analysis and metrics used to evaluate the accuracy of the presented method. This reviewer has provided the following recommendations to improve this manuscript.

Page 2, Line 41-43: Please provide appropriate references for each of the applications of TLS systems within civil engineering domain.

Page 2, Line 45: Not having "drawings" is not the ONLY reason to use TLS. Maybe in the case of generating as-builts, this statement makes sense. However, this study is all about condition assessment where the major benefit of TLS is in the fact that this is a fast and fairly accurate remote sensing.

Page 2, Line 50: There have been some interesting works on deformation and damage analysis using raw point cloud data, such as:

Jafari, B., Khaloo, A., & Lattanzi, D. (2017). Deformation tracking in 3D point clouds via statistical sampling of direct cloud-to-cloud distances. Journal of Nondestructive Evaluation, 36(4), 65.

Dai, K., Li, A., Zhang, H., Chen, S. E., & Pan, Y. (2018). Surface damage quantification of the post-earthquake building based on terrestrial laser scan data. Structural Control and Health Monitoring, e2210.

Dawood, T., Zhu, Z., & Zayed, T. (2017). Computer vision–based model for moisture marks detection and recognition in subway networks. Journal of Computing in Civil Engineering, 32(2), 04017079.

Law, D. W., Silcock, D., & Holden, L. (2018). Terrestrial laser scanner assessment of deteriorating concrete structures. Structural Control and Health Monitoring, 25(5), e2156.

Khaloo, A., & Lattanzi, D. (2019). Automatic Detection of Structural Deficiencies Using 4D Hue-Assisted Analysis of Color Point Clouds. In Dynamics of Civil Structures, Volume 2(pp. 197-205). Springer, Cham.

Please consider adding them to your literature review.

Page 2, Line 52: It is the first time to use"NURBS" in the paper. Please define the abbreviation, as NURBS stands for "Non-Uniform Rational B-Splines"

Page 2, Line 64: Please provide some references on using neural networks for damage detection.

Page 2, Line 68: Please add "." at the end of the sentence.

Page 3, Line 102: Using TLS does NOT allow to "understand" complex objects, it helps us with easy data collection to generate a 3D virtual model of the physical asset. Also, define the term "point clouds" as a set of points in 3D space.

Page 4, Line 118: Define abbreviations such as DB or DF upon their first appearance in the text.

Page 4, Line 138: Please provide a citation for Tamminen method.

Page 11, Section 4.2: It seems that the authors have decided to use the intensity data collected by the TLS for their crack detection method. However, TLS also provides RGB data of the surface in addition to intensity. Although the RGB data tends to be more sensitive, the authors should at least add a sensitivity study to compare using RGB with intensity. Please check the study by Hou et al. (2017)

Hou, T. C., Liu, J. W., & Liu, Y. W. (2017). Algorithmic clustering of LiDAR point cloud data for textural damage identifications of structural elements. Measurement, 108, 77-90.

Page 12, Line 288: Please add the citation for the Octovis to the paper

Hornung, A., Wurm, K. M., Bennewitz, M., Stachniss, C., & Burgard, W. (2013). OctoMap: An efficient probabilistic 3D mapping framework based on octrees. Autonomous Robots, 34(3), 189-206.

Page 12, Line 290: What is the format of the laser scanned data? Is it PLY, LAS, LAZ, or any other point cloud format? Please indicate it

Page 13, Figure 12: Please provide the Octree Level used for the point cloud compression in each case.

Page 13, Line 298: Since authors made a snapshot of the original point cloud model, why they used 3026 x 1620 pixels, while they could zoom-in and even make a better resolution image. Please elaborate.

Page 13: It is standard practice to measure point cloud compression performance in bits per point (bpp) and the loss of quality by the peak signal to noise ratio (PSNR). Please provide these analyses in your work. You can find more about bpp and PSNR in papers such as Schnabel, R., & Klein, R. (2006). Octree-based Point-Cloud Compression. Spbg, 6, 111-120.

Page 14: There is an absolute need to add a much better quantitative analysis to test the performance of the presented algorithm. The authors are highly encouraged to use standard metrics such as Precision, Recall, and F1-score to evaluate their method and compare it against the state-of-the-art.

Page 14, Line 354: Please provide a proper citation for the Taleb's Method.

Page 14: The authors are highly encouraged to compare their algorithm with more advanced methods from the literature, to provide better insights into the strengths of their method.

Author Response

1. Page 2, Line 41-43: Please provide appropriate references for each of the applications of TLS systems within civil engineering domain.

As the other reviewer also pointed out the lack of literature review, introduction has been considerable improved by using proper references of various applications of TSL.

2. Page 2, Line 45: Not having "drawings" is not the ONLY reason to use TLS. Maybe in the case of generating as-builts, this statement makes sense. However, this study is all about condition assessment where the major benefit of TLS is in the fact that this is a fast and fairly accurate remote sensing.

As the reviewer comments, the specific purpose of TSL such as "drawings" has been removed and the general purpose of TSL has been added using the fact that "TSL is a fast and fairly accurate remote sensing"

3. Page 2, Line 50: There have been some interesting works on deformation and damage analysis using raw point cloud data, such as:

Jafari, B., Khaloo, A., & Lattanzi, D. (2017). Deformation tracking in 3D point clouds via statistical sampling of direct cloud-to-cloud distances. Journal of Nondestructive Evaluation, 36(4), 65.

Dai, K., Li, A., Zhang, H., Chen, S. E., & Pan, Y. (2018). Surface damage quantification of the post-earthquake building based on terrestrial laser scan data. Structural Control and Health Monitoring, e2210.

Dawood, T., Zhu, Z., & Zayed, T. (2017). Computer vision–based model for moisture marks detection and recognition in subway networks. Journal of Computing in Civil Engineering, 32(2), 04017079.

Law, D. W., Silcock, D., & Holden, L. (2018). Terrestrial laser scanner assessment of deteriorating concrete structures. Structural Control and Health Monitoring, 25(5), e2156.

Khaloo, A., & Lattanzi, D. (2019). Automatic Detection of Structural Deficiencies Using 4D Hue-Assisted Analysis of Color Point Clouds. In Dynamics of Civil Structures, Volume 2(pp. 197-205). Springer, Cham.

Please consider adding them to your literature review.

As the reviewer comments, the introduction has been considerably improved using the recommend references.

4. Page 2, Line 52: It is the first time to use "NURBS" in the paper. Please define the abbreviation, as NURBS stands for "Non-Uniform Rational B-Splines"

The term "NURBS" has been defined as ""Non-Uniform Rational B-Splines" in the manuscript.

5. Page 2, Line 64: Please provide some references on using neural networks for damage detection.

Line 64 has been changed with proper references as

Moon and Kim [31] used neural network to detect cracks from images, and Kawamura et al. [32] used genetic algorithm to extract crack patterns from digital images.

[31: Neural network]

Moon, H., & Kim, J. (2011). Intelligent crack detecting algorithm on the concrete crack image using neural network. Proceedings of the 28th ISARC, 1461-1467.

[32: Genetic Algorithm]

Kawamura K., Miyamoto A., Nakamura H., Sato R.: Proposal of a crack pattern extraction method from digital images using an interactive genetic algorithm. Proc. Japan Soc. Civil Eng. 742, 115–131 (2003)

6. Page 2, Line 68: Please add "." at the end of the sentence.

The comma “.” is added at the end of the sentence as the reviewer points out.

7. Page 3, Line 102: Using TLS does NOT allow to "understand" complex objects, it helps us with easy data collection to generate a 3D virtual model of the physical asset. Also, define the term "point clouds" as a set of points in 3D space.

As the reviewer suggests, the sentence "TSL allows to understand ~" has been modified as the following.

Terrestrial laser scanning is a technology that helps users to get easy data collection in generating a 3D virtual model of the physical asset using point clouds as a set of points in 3D space.

8. Page 4, Line 118: Define abbreviations such as DB or DF upon their first appearance in the text.

The first appeared abbreviations such as DB and DF have been defined as the full name like database (DB), depth first (DF)

9. Page 4, Line 138: Please provide a citation for Tamminen method.

The proper reference for Tamminen method has been added.

10. Page 11, Section 4.2: It seems that the authors have decided to use the intensity data collected by the TLS for their crack detection method. However, TLS also provides RGB data of the surface in addition to intensity. Although the RGB data tends to be more sensitive, the authors should at least add a sensitivity study to compare using RGB with intensity. Please check the study by Hou et al. (2017)

The reference of Hou et al. (2017) used clustering algorithms for both intensity and RGB data for three types of damages: metal rusting, tile spall off, and water staining on finished wall. The clustering algorithm worked better with the RGB data on metal rusting and tile spall off, since these two damages types makes significant change on the colors. Meanwhile, the clustering algorithm worked better with the intensity data on water staining, since this damage does not involve serious color change. (See Table 5 of Hou et al. (2017)).

We have also tried our method on both intensity data with RGB collected by the TLS, and any remarkable advance has not been observed. That’s why we aim to detect cracks on the concrete wall, in which no significant color change is involved. This is very similar to the water staining case of Hou et al. (2017).

But the reference of Hou et al. (2017) is very relevant to our study, and we will progress our study for other types of structures, such as tiled buildings and metal structures, that has colorful background. We really appreciate the reviewer’s suggestion.

11. Page 12, Line 288: Please add the citation for the Octovis to the paper

The following reference of Octovis has been added.

Hornung, A., Wurm, K. M., Bennewitz, M., Stachniss, C., & Burgard, W. (2013). OctoMap: An efficient probabilistic 3D mapping framework based on octrees. Autonomous Robots, 34(3), 189-206.

12. Page 12, Line 290: What is the format of the laser scanned data? Is it PLY, LAS, LAZ, or any other point cloud format? Please indicate it

The format of the laser scanned data has been indicated as "PTS" in the manuscript.

13. Page 13, Figure 12: Please provide the Octree Level used for the point cloud compression in each case.

The octree level in the all the cases has been informed in the Table 3.

14. Page 13, Line 298: Since authors made a snapshot of the original point cloud model, why they used 3026 x 1620 pixels, while they could zoom-in and even make a better resolution image. Please elaborate.

The original snapshot resolution was 1513 x 810 pixels, but a 3026 x 1620 pixel image with the doubled resolution was applied for precise crack detection. Also, this contents has been revised in the manuscript.

15. Page 13: It is standard practice to measure point cloud compression performance in bits per point (bpp) and the loss of quality by the peak signal to noise ratio (PSNR). Please provide these analyses in your work. You can find more about bpp and PSNR in papers such as Schnabel, R., & Klein, R. (2006). Octree-based Point-Cloud Compression. Spbg, 6, 111-120.

The main point of our study is that main cracks can be effectively identified by applying the K-means clustering on the image extracted from the octree technique of TSL data. In this regard, we did not examine the cloud quality such as bpp and PSNR deeply, but will review it in the next study.

16. Page 14: There is an absolute need to add a much better quantitative analysis to test the performance of the presented algorithm. The authors are highly encouraged to use standard metrics such as Precision, Recall, and F1-score to evaluate their method and compare it against the state-of-the-art.

17. Page 14, Line 354: Please provide a proper citation for the Talab' Method.

The reference of Talab's method has been added.

18. Page 14: The authors are highly encouraged to compare their algorithm with more advanced methods from the literature, to provide better insights into the strengths of their method.

For reasonable comparative analysis with this study, we referred to 2016 Utah Transportation Center Report "Automatic Surface Crack Detection in Concrete Structures using Otsu thresholding and Morphological Operation". It reviewed various recent imaging processes, and reported that Talab's method is one of the most applicable and effective imaging processes in the field.

Also, we could not find other methods other than K-means clustering to efficiently distinguish cracks from sediments, joints, and efflorescence. We do not mean that this study outperforms other imaging process methods, but that the K-means clustering can be an effective way to distinguish cracks from others especially in the laser scanning image.

State of the art report

Sattar Dorafshan, Marc Maguire, Xiaojun Qi, Automatic Surface Crack Detection in Concrete Structures using Otsu thresholding and Morphological Operation, Utah Transportation Center Report 01-2016

References of the report

Ammouchea, A., D. Breysseb, H. Hornaina, O. Didryd, and J. Marchandc. 2000. "A new image analysis technique for the quantitative assessment of microcracks in cement-based materials." Cement and Concrete Research 25-35.

Aoki, S., and T. Nagao. 1999. "Automatic construction of treestructural image transformation." International Conference on Image Processing. Washington DC: IEEE Computer Society .

Ikhlas Abdel-Qader, P.E, P.E., M.ASCE Osama Abudayyeh, and and Michael E. Kelly. 2003. "Analysis of Edge-Detection Techniques for Crack Identification in Bridges." JOURNAL OF COMPUTING IN CIVIL ENGINEERING 255-263.

Kim, Jong-Woo, Sung-Bae Kim, Jeong-Cheon Park, and Jin-Won Nam. 2015. "Development of Crack Detection System with Unmanned Aerial Vehicles and Digital Image Processing." Advances in structural engineering and mechanics (ASEM15). Inchoen, Korea: I-ASEM.

Kittler, J., and J. Illingworth. 1986. "Minimum error thresholding." Pattern Recognition 41-47.

Kittler, J., R. Marik, M. Mirmehdi, M. Petrou, and J. Song. 1994. DETECTION OF DEFECTS IN COLOUR TEXTURE SURFACES. Guildford: University of Surrey.

Litorowicz, Agnieszka. 2006. "Identification and quantification of cracks in concrete by optical fluorescent microscopy." Cement and Concrete Research 1508-1515

Matsumoto, Masato, Koji Mitani, and F. Necati Catbas. 2013. BRIDGE ASSESSMENT METHODS USING IMAGE PROCESSING AND INFRARED THERMOGRAPHY TECHNOLOGY. Orlando: University of central Florida.

Moon, Hyeong-Gyeong, and Jung-Hoon Kim. 2011. "Inteligent Crack Detecting Algorithm On The Concrete Crack Image Using Neural Network." 28th ISARC. Seoul. 1461-1467.

Nishikawa, Takafumi, Junji Yoshida, Toshiyuki Sugiyama, and Yozo Fujino. 2012. "Concrete Crack Detection by Multiple Sequential Image Filtering." Computer-Aided Civil and Infrastructure Engineering 29-47.

Oliveira, Henrique, and Paulo L. Correia. 2013. "Automatic Road Crack Detection and Characterization." IEEE Transactions on Intelligent Transportation Systems 155-168.

Otsu, N. 1979. "A Threshold Selection Method from Gray-Level Histograms." IEEE Transactions on Systems, Man, and Cybernetics 62-66.

Pereira, Fábio Celestino, and Carlos Eduardo Pereira. 2015. "Embedded Image Processing Systems for Automatic Recognition of Cracks using UAVs." IFAC. PortoAlegre: ELSEVIER. 016-021.

Rimkus, Arvydas, Askoldas Podviezko, and Viktor Gribniak. 2015. "Processing digital images for crack localization in reinforced concrete members." Procedia Engineering 239-243.

Sankarasrinivasana, S., E. Balasubramaniana, K. Karthika, U. Chandrasekarb, and Rishi Gupta. 2015. "Health Monitoring of Civil Structures with Integrated UAV and Image Processing System." Procedia Computer Science 508-515.

Talab, Ahmed Mahgoub Ahmed, Zhangcan Huanga, Xi Fan, and HaiMing Liu. 2015. "Detection crack in image using Otsu method and multiple filtering inimage processing techniques." Optik 1-4.

Wang, Hui, Zhang Chen, and Lijun Sun. 2013. "Image Preprocessing Methods to Identify Micro-cracks of Road Pavement." Optics and Photonics Journal 99-102.

Yamaguchi, Tomoyuki, and Shuji Hashimoto. 2010. "Image-Based Crack Detection for Real Concrete Surfaces images using percolation-based image processing." Machine Vision and Applications (Waseda University) 797-809.

Yamaguchi, Tomoyuki, Shingo Nakamura, Ryo Saegusa, and Shuji Hashimoto. 2008. "Image-Based Crack Detection for Real Concrete Surfaces." TRANSACTIONS ON ELECTRICAL AND ELECTRONIC ENGINEERING 128-135.

Zhang, Wenyu, Zhenjiang Zhang, Dapeng Qi, and Yun Liu. 2014. "Automatic Crack Detection and Classification Method for Subway Tunnel Safety Monitoring." Sensors 19307-19328.

Zheng, Paul, and Cristopher Moen. 2014. CRACK DETECTION AND MEASUREMENT UTILIZING IMAGE-BASED RECONSTRUCTION. Blacksburg: VIRGINIA POLYTECHNIC INSTITUTE AND STATE UNIVERSITY.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

While the authors have addressed many.of the issues the fact that only 1 case is presented remains a problem. A case is a physical object not a change in resolution or how the data are collected. This remains one of two major outstanding problems in the paper, which would then allow the results of the current cases to be moved into a discussion section; the absence of which is the other current deficit of the paper.

Author Response

We sincerely thank you for the constructive comments.

Review 1

As the reviewer comments, the results for only one case seems to be unreliable for the verification. Therefore, we added the results of a clear crack case (with some stains) which has been generally used for the imaging process of crack detection in the Figure 15 on the page 18. Unlike typical image processing methods, we want to inform that the K-means clustering is very effective in distinguishing main cracks from others such as stain, sediment, and effloresce. Also, the proposed method is available for the general purpose.

Reviewer 2 Report

Based on the revisions made by the authors, this paper is now ready for publication.