Geometric Quality Assessment of Prefabricated Steel Box Girder Components Using 3D Laser Scanning and Building Information Model

Round 1

Reviewer 1 Report (Previous Reviewer 1)

This is resubmitted manuscript and the authors improved it significantly. They addressed all the concerns from my previous review and manuscript deserves to be published as a paper in the journal Remote Sensing (MDPI).

Author Response

Thank you for your valuable opinions! This article has checked and corrected some grammatical errors again. And we have asked a professor, who is well established expert, to polish our paper.  Please see if the revised version met the English presentation standard.

Author Response File: Author Response.docx

Reviewer 2 Report (New Reviewer)

The authors proposed a non-contact geometric quality assessment technique based on 3D laser scanning to assess the locations and dimensions of the SBG components. The proposed technique was validated through field test results. Looks like the manuscript has been revised based on previous review? The article is very well-written and has a clear structure. The reviewer has two minor comments as following:

1.     The abstract mentioned that “the developed inspection technique could assess the geometric quality of prefabricated SBG components in a more accurate and efficient manner”, but the field test results did not mention anything related to the efficiency of the method compared with manual ones, e.g. required time, speed, etc. It would be good to see some discussion/results on this aspect.

2.     The titles of Section 2.2-2.4 are not very clear. One suggestion is to specify more of the title, e.g. Object recognition in point cloud? Object recognition of bridge components? Location recognition of/in xxx? Quality assurance and control of/in xxx?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

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

Comment: The manuscript shows a research on connection of point cloud analysis and BIM for quality control of build construction element. The introduction and literature review are of very good quality as well as the methodology of the research. On the other side, I have major concerns on the experiment and feasibility part of the research that are defined below with other minor remarks. Also, the manuscript does not follow editing instructions defined in journal site (https://www.mdpi.com/journal/remotesensing/instructions) and one figure is of very low qulity. Below you can find my concerns with major remarks described in bold.

Remarks:

1. References are defined with author-year which is not in accordance with instructions for authors.

2. There are no line numbers, so it is a bit hard to point out the position in the manuscript where change should be made. This is also not in accordance with instructions for authors.

3. Figure 2. is of very low quality. There are unreadable descriptions on both parts of the figure.

4. 4th level heading must be unnumbered according to the instructions for authors (3.2.1.1. Normal estimation, 3.2.1.2. Region growing segmentation).

5. Tabes 3. and 4. do not specify the measurement units.

6. My main concern is regarding experiment provision, description and analysis. It is not straightforward how you manage the accuracy and precision of different tools and how you describe and compare the results. You wrote that the accuracy of the scanner is 2 mm and there is no information on precision. On the other hand you measured characteristic points with steel tape and there are no information on its precision and accuracy (deviation of 1 mm is what precisely?). And then you compare those two somewhere up to the 1 decimal space of the millimeter, and somewhere up to 2 decimal spaces of the millimeter, as well as you defined RMSE with 3 decimal spaces of the millimeter which is quite odd because I am quite sure that you did not measure the constructed element up to 1 micrometer (maybe I am wrong, convince me). To clear this, please follow the standard procedures for managing measurements and comparison between different measuring tools. Below is one example where it is unclear how you manage the measurement and describe the results, please redo it in all the parts of the analysis.

"...Figure 12a shows the error analysis histogram for T rib positions, the maximum and minimum absolute differences were 12.5 mm and 0.18 mm, respectively, and the RMSE was 9.018 mm..."

 

Reviewer 2 Report

I only reccomend, from a methodological point of view, to add some more description of the geometric genesis and configuration of the analyzed structures. In this way, the bibliographic references can be enriched with citations of studies about the geometric surfaces analysis. Finally, refering to the section of real SBG on the prefabrication site, I suggest to add an image of the entire architectural/engineering case study.

Reviewer 3 Report

This paper proposes an automatic quality assessment algorithm for prefabricated steel box girders by using 3D point cloud processing and BIM models. The paper mainly proposes point cloud segmentation algorithms of T-shaped and I-shaped steel members attached to a girder that is robust against measurement noises. The proposed algorithm was validated in a real girder structure. 

 

Although the algorithms the author proposed might be effective to efficient quality inspections of steep box girder components, there are still many improvements that need to be made to the manuscript before the paper can be considered for acceptance: in particular, the clarity of the illustrations, and the mathematical precision of the formulas.

 

Comments for the major revisions:

1. Figure 2: Currently, there is no explicit indication of where the I-beams and T-ribs are in the SBG. The authors should clearly indicate them as early as possible in the paper. 

2. Section 3.1: The authors should not discuss the concrete value (8) of the parameter MinPts of DBSCAN in this theory section because the generality of the algorithm might be loose and it should be changed according to the scanner point cloud density. Instead, the authors should only provide the general selection guideline of this parameter. For example, the relation of the parameter to the average point density nearby.

3. Equation(1): The authors should clearly state how the robust distance (Eq(1)) was used for the normal vector calculation at each point. Moreover, the definition of a covariance matrix estimated by MCD does not need the inverse symbol. Please check. 

4. Equation (2): This equation is trivial, and should be removed.

5. Figure 4: The viewpoints and axis scales of Figure 4(a) (b) and (c) should match.

6. Equation(3): e_i does not appear in the right hand side. The authors should more state what e_i means. 

7. Equation(4): The authors should clearly state that the top boundary detection using mean shift algorithm was performed in 2-dimenionally (on the projected plane) or 3-dimensionally.

8. Section 3.3.1: Is this section really needed? The substantial processing stated in the section is very trivial (just taking a set difference between two point clouds.)  

9. Section 4.1.2: The detail of the RANSAC-based coarse registration algorithm was not stated and should be stated.

10. Figure 6(d):  In this example, the shape of the BIM model does not match with the measured point cloud (in the root portions of T-ribs). The authors should state how they handled these unmatched portions in their registration algorithm.  

11. Section 4.2.1.1: It was unclear whether the front side plane and back side plane of T-ribs could be measured or not and whether they need to extract both planes for the inspection or not.   

12. Figure 8: The viewpoints and axis scales of Figure 8(a) and (c) should match.

13. Section 4.2.1.3: The authors should state why they should perform the complex process in this section to calculate the inspection points and why they could not just compute the intersection points of the detected three planes.

14. Figure 8: The viewpoints and axis scales of Figure 8(a) to (e) should match.

15. Equation (7): The authors should clearly indicate how they use the covariance matrix to calculate an OBB. 

16. Section 4.2.1.4:  The distance between adjacent T ribs would change depending on the evaluation point the authors selected (close to the rib root or tip). The authors should sate how they determined the distance-examination point on the plane. 

17. Equation (8): The definition of n1 and n2 should be provided. The function “acosd” should be defined. 

18. Figure 11: Changes in the point cloud before and after denoising should be shown to confirm the effectivity of the proposed DBSCAN algorithm.

19. Figure 11(f): Was the upper boundary of this figure extracted by the original mean shift algorithm the author proposed in 3.2.2 or by using the different algorithm? What does “improved” edge detection algorithm mean? 

 

Comments for the minor revisions:

20. Page 2, Line 13 from the bottom: “geometric semantic information” is too abstract to understand. The author should modify it to more plain and concrete expressions.

21. References: Since there are still other related studies on the normal-based region growing for laser-scanned point cloud segmentation, the authors should consider adding them to the reference list. Especially, in the main text, the authors did not state any comment on the following reference (Khaloo 2017) that is indicated in their reference list.

* Khaloo, A., D. Lattanzi “ Robust normal estimation and region growing segmentation of infrastructure 3D point cloud models,” Advanced Engineering Informatics, 34, 2017, 1-16, https://doi.org/10.1016/j.aei.2017.07.002.

* “As-built modeling of piping system from terrestrial laser-scanned point clouds using normal-based region growing,” J. Computational Design and Engineering, 1(1):13-26, 2014. https://doi.org/10.7315/JCDE.2014.002

22. Figure 1: Some typos appeared in the figure: “Create” “Region grow”

23. Figure 2: The red texts in the figure are unclear and not visible. The authors should correct them.

24. Section 3.3.2: The statement “the noise caused by occlusion” should be reconsidered, because an occlusion only causes missing portions on the point cloud. 

25. Section 4.1.3: The calculation method of the area on a point cloud should be provided. 

26. Section 5: The photo and rough size of the SBG should be provided. 

27. Section 5.1: The size and average point density of the scanned point cloud should be provided.

 

Comments for author File: Comments.pdf