Enhanced Point Cloud Slicing Method for Volume Calculation of Large Irregular Bodies: Validation in Open-Pit Mining

Round 1

Reviewer 1 Report (Previous Reviewer 1)

After careful review of the authors' responses and revisions, I believe that the authors have revised the manuscript sufficiently to be considered for publication in Remote Sensing.

Author Response

Dear reviewer

I hope this letter finds you well. I genuinely appreciate your thorough review of our manuscript and your validation of our efforts. I extend my best wishes for your well-being .

Tianyi Wang

29 August 2023

Reviewer 2 Report (New Reviewer)

In this paper, an enhanced point cloud slicing method is proposed. The process involves slicing the point cloud data at regular intervals in predetermined directions. The slices with multi-contour boundaries are then segmented using the Euclidean clustering method, and the concave hull algorithm is employed to extract the contour polygons of each slice. Finally, the volume is calculated by multiplying the area of each polygon by the spacing and summing up the products to obtain the final volume. To validate the effectiveness and accuracy of our method, we utilize the volume determined by the model as the ground truth and compare the errors generated by both the traditional slicing method and the approach proposed in this paper. The enhanced point cloud slicing method proposed in this paper has good error accuracy for open-pit mine volume measurement and has certain application value, but there may still be the following problems :

In Line 301 : Euclidean clustering segmentation results are easily affected by clustering thresholds. What is the basis for using this clustering threshold calculation formula in this paper ? It is necessary to further explain and analyze the formula.

In Line 373 : Taking the first set of experimental data as an example, why the final results are shown in Figure 7 ( b ) and ( l ), please check it.

In Section 3.3.2 : In the proposed method, the Euclidean clustering method is introduced to segment the slice results. As an important part of the method in this paper, it is better to provide a schematic diagram of the clustering results of some sample data. Like using same color points, it will be easy to understand the article.

In Line 508 : In the process of the author 's introduction, the Euclidean clustering method is first used for segmentation, and then the traditional slicing method is used to extract and sort the contour boundary. The Euclidean clustering method should be a part of the slicing method according to the previous article. Is the Euclidean clustering method mentioned here a preprocessing process or the traditional slicing method also using the Euclidean clustering method for classification and then using the two-way nearest point search method ? Please  clearly pointed out what improvements have been made in this paper for the traditional slicing method.

In Section 3.4.4 : The article only calculated the algorithm operation time, but did not mention the operation environment requirements.

As shown in Figure 11 of the article, the error caused by the traditional slicing method is mostly due to the setting of the starting and ending point. Compared with the traditional method of two-way nearest point search method, this paper used the concave algorithm to better extract the boundary. However, this method has been applied to point cloud boundary extraction in some earlier articles. For example, the author cited in the literature:

Yan Z, Liu R, Cheng L, et al. A concave hull methodology for calculating the crown volume of individual trees based on vehicle-borne LiDAR data[J]. Remote sensing, 2019, 11(6): 623.

When measuring the volume of large-volume irregular bodies, the volume measurement method in this paper only used the polygon area multiplied by the spacing, and then used the product summation as the final volume, ignoring the area change of the irregular body within the slice thickness. Obviously, a single fixed spacing cannot accurately express the volume of the irregular body, and the area located in the middle of the two slices will be ignored in the volume measurement process, resulting in errors. Although the author explains this deficiency in DIscussion, the use of adaptive thickness still cannot solve the error problem of calculating the polygon area multiplied by the spacing as the volume.

In this paper, the idea of using airborne lidar technology to measure the volume of open-pit mine is feasible. In the preprocessing process, the non-overlapping area is determined by the adjacent point search, and good results are also achieved. However, the follow-up method only applies the combination of clustering and concave hull method to the open-pit mine volume measurement. This method has been applied to model extraction by some scholars in recently years, which lacks certain innovation. It is suggested to deepen the innovation point. In the process of result comparison, only a traditional slicing method is compared, which is not very convincing. It is suggested to add several comparison models for further verification.

At last, it will be better provide the basis for calculating the results, such as publicly part of the source code, this will enhance the persuasiveness of this article and facilitate further understanding of the content of this article.

Author Response

Dear reviewer

Please see the attachment.

Tianyi Wang

Author Response File: Author Response.pdf

Reviewer 3 Report (New Reviewer)

This paper aims to propose an enhanced point cloud slicing method that combines Euclidean clustering with the concave hull algorithm. Experimental results demonstrate that the proposed point cloud volume calculation method achieves an average relative error of 1.17%. The proposed workflow is generally feasible, but the scientific expression and logical organization are not good, and thus need to be carefully revised and improved. There are some detailed comments as below:

(1)   When the point cloud slice method is used to calculate irregular volumes, Euclidean clustering for boundary segmentation, and concave packet algorithm for boundary sorting have all been proposed in the existing literature. Where is the improvement of them in this paper? The innovation is only the combination of use?

(2)   The abstract needs to be further streamlined to ensure conciseness and accuracy of expression.

(3)   The figures are poorly expressive and lack corresponding legends or descriptive language, including Figure 1, flowchart (Figure 2), Figure 3, Figure 6, Figure 7, and Figure 8 (check other figures at the same time).

(4)   How is the point cloud in the multi-temporal open-mining area (i.e., phase 1 and phase 2) well registered?

(5)   The existing volume calculation methods for irregular objects should include three types, namely, voxel method, slice method, and internal model method. The paper does not appear to describe related work on voxel methods and does not include some newer literature on irregular volume calculation methods (such as Ref. 1, authors of other recent literature should ensure comprehensive searches and necessary citations).

(6)   Ref. 2 (reference 22 of this manuscript) mentioned that the two methods based on Euclidean clustering segmentation method and polygon splitting and reorganization can divide the boundary However, in complex boundary segmentation (the distance between the boundaries is relatively close, and there is noise point interference), the segmentation results of the polygon splitting and reorganization method are significantly better than the Euclidean clustering method (the volume error is less than 0.1 %, which is better than the 1.17% given in this paper). Therefore, the authors should try to increase the segmentation comparison with the polygon splitting and reassembly method to further illustrate the advantages of the proposed method.

(7)   The formula is not universally expressive, and the expression of the table is not good, such as formula 5 and table 2.

(8)   The presentation organization of the research methods, research results and discussion sections is not good, and the author is suggested to reorganize and improve. Two aspects can be considered for improvement: (1) accuracy of expression; (2) simplicity of expression.

 

Ref. 1：Cheng X. J., Xiong X. X., Yang Z. X. et al., 2019. Cavern Capacity Calculation Using Terrestrial Lidar. Laser & Optoelectronics Progress, 56(23): 231201.

Ref. 2：Liu, J. J., and H. J. Li. "Volume measurement of irregular objects based on improved point cloud slicing method." Acta Optica Sinica 41.23 (2021): 2312003.

Moderate editing of English language required

Author Response

Dear reviewer

Please see the attachment.

Tianyi Wang

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report (New Reviewer)

In the latest revised manuscript, the author has revised part of the original manuscript and made some improvements, but the changes are not great. The author did not give a specific description of the algorithm innovation, and it is not enough to publish the combination scheme that only expands the application scenarios, because it is obvious from Table 6 that the improvement in computing efficiency is actually not large, but it is indeed compared with the most traditional The scheme improves accuracy. After comprehensive consideration, it is recommended to add necessary comparative experiments and innovative detailed statements (note that the authors of these comments should also reply). A review of some details follows:

 

(1) Regardless of the reason, previous research on the use of voxel methods for irregular volume calculations should be reflected in the introduction, even if they are not completely correlated and matched, which is a basic requirement for scientific research.

(2) Please add a comparison experiment of polygon segmentation and reorganization, qualitative description and explanation are not enough to explain the advantages of the proposed scheme.

(3) The overall language description/diagram drawing/formula summary of the article still needs to be improved (suggestion).

A review and revision of the language of the full text may be considered

Author Response

Dear reviewer

Please see the attachment.

Tianyi Wang

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

To address the limitations of traditional point cloud slicing methods, which fail to distinguish between multi-contour boundaries and ex tract anomalous boundary polygons when calculating excavation amounts in open-pit mines, they propose an enhanced point cloud slicing method that combines Euclidean clustering with the concave hull algorithm. Firstly, the point cloud data from the open-pit mine are sliced at regular intervals along a predetermined direction sequence.

The manuscript is innovative, but there are significant problems with both the quality of the presentation and the English style of the article, and only careful revisions will make it possible to be considered for acceptance by Remote Sensing.

1.The abstract of the manuscript is so cumbersome that it makes reviewers and readers feel that it is not a research article but a science article. The abstract should simply give the motivation for the study and the research methods and landmark results. This should be completely rewritten.

2.At the end of the introductory section, it is sufficient to keep the introduction to the method or to keep the description of the contribution of the manuscript. It is not necessary to repeat the description of the contributions (highlights) of the paper.

3.The resolution of the images throughout the manuscript is so poor that it makes the overall quality of the presentation of the article a concern, and it is recommended that high resolution images be repositioned.

4.Lines442-458，Please describe the formulae according to the specification. Please refer to the template provided by the journal for details.

5.A number of comparisons in the experimental section show from the results that the work is adequate and good results are obtained. It is not important, only my query, that the paper mentions that the proposed point cloud calculation method is applicable to a wide range of detection regimes including UAV and Lidar. As far as I know, the detection regimes of these instruments are not all the same, for example Lidar contains different temporal and spatial resolutions. Do these factors not have an impact when it comes to the actual calculations using the proposed method. In other words, if the Lidar is not an imaging Lidar but a ranging Lidar, will the proposed calculation method still give the same results? These need to be clarified or verified.

 

To put it mildly, the English style and presentation of this manuscript does not, in my opinion, meet the requirements of the journal (either in terms of grammar or style of writing). The manuscript should be rewritten or a native English-speaking researcher should be involved.

Reviewer 2 Report

review comments on "Calculating Excavation Quantity in an Open-Pit Mine Using the Enhanced Point Cloud Slicing Method"

 

This article presents an enhanced slicing method for computing open pit excavation volume. The main problem with the article is that the reference method is not clearly presented. 

Allegedly, there is some method that is not able to handle contours with multiple closing lines at the same height. This sounds strange, as the contours are widely used in different tasks and thoroughly studied already a hundred years ago. It is unfortunate, if such method has been used during the last 20 years. Algorithm for comparison method is missing. As was mentioned in the introduction, many improvements have already been suggested in other studies. In my opinion, comparison with these studies is relevant and needed - also comparison of results and time consumption of triangulation- and voxelbased methods. I recommend this article to be rejected. There obviously has been a lot of work involved in developing this method, but it is not scientifically interesting. The work consists of combination of elements that are well known and used - if not in specifically open-pit mines but elsewhere. 

Currently, the solutions for volume determination are limited to slicing methods, computational cost is one of the reasons given for abandoning other solutions. In my opinion, you need to give specific reasons why some solutions are not considered. "triangulation is slow" is not specific enough. And for the family of slicing based solutions, you need more details and most importantly, comparison with other enhanced methods and very clear explanation of all steps and differences between methods - preferably open code.

Ideally, the reference would be TLS or ALS - different point cloud altogether.

 

I am sure you can use the point cloud data in an article that compares different methods of volume computation. Including the presented one. 

 

Computational cost considerations in open-pit mine excavation volume computations are interesting, is there any need to be quick with the computation? Answer is no. There may well be many volume computing tasks in which near real time is needed, but not this. You could discuss this and perhaps set some time limit inside which the computation should be ready, but there is no harm if one method is a couple of minutes slower than another in total. What takes time is careful planning and operation of data collection. I claim that in volume, most of the error is caused by deformation from SLAM problems, registration errors and other sources not related to the volume computation method. As to the volume computation, your reference method is the way I suppose a private company hired to compute the excavation volume would handle the task. So where is the presented method needed and who needs it?  You maybe can answer this - you perhaps need a method that works on 80's computer or some other specification - you need to tell it. That way seemingly odd selections can be relevant.

 

Note on registration of point clouds from different times: I got the understanding that the point cloud captures from different times were individual projects and the registration was made by you. I think it is quite essential, that the data capture is planned so, that there are enough permanent objects, that are used in coregistration. It is essential to know if this was the case here or not.

 

Specific comments

lines 58 -59: "The current methods for calculating the volume of irregular point clouds are mainly divided into the inverse modeling method and the point cloud slicing method [11]." The reference is 10 years old and discusses tree canopy volumes. It is hardly the current state of the art in mining related remote sensing.2

line 114: please clarify and add reference - what is meant by European clustering and concave hull algorithms.

lines 171 - 173: the used terminology is not consistent, P1 and P2 are the names of point clouds that were captured on different times. On page 5 lines 210 - 211 "the first period point cloud to red and the second period point cloud to green". On page 9 line 339 "first phase point cloud" and "second phase point cloud" are used.

lines 199 - 223: This section is very difficult to understand - you have a method for determining if the change is real excavation or other change. Excavation being disappearance of material in at least some border quantity and other change less than border disappearance or addition of material. The color system is not easy and the inequalities are confusing.

line 341: How do you determine a "scatter point" In my opinion "point" is better term. Unless you specifically want to make a distinction between lidar and photogrammetry.

line 231: Intuitively, the computation of contours using only one height slice at a time seems complicated versus using gridded solution. It is very unfortunate, that the cited work is in Chinese. It would be interesting to know if the triangulation or grid solutions manage the areas where slicing is having troubles.

lines 337 - 339:  "the first phase data contains 28068437 scattered points and the second phase data contains 14065581 scattered points; furthermore, the scope of the two phases’ 338 point clouds is the same." Please be more specific. Was the ground sample distance the same in both sessions or larger in the later one? Does "same scope" mean that you have delineated the same area from both full point clouds?

Figure 9: The scale is missing from the plots. This makes it difficult to read the contours - one is wondering if there are floating objects because the image box is the same size and there is no scale on either of the axes.

line 448: "In this paper, we use the Shoelace Theorem described in Section 2.1 to calculate the..." Please use the term shoelace theorem in section 2.1. and add reference.

lines 622 - 627: This result could be computed from artificial data and if the method was novel in mathematics, I would encourage you to do computations and comparisons with other methods in artificial data, that could visually clarify the methods and their differences. 

lines 629 - 631: "The enhanced point cloud slicing method can effectively solve the difficult problem of multi-contour boundary when the traditional slicing method is applied to the calculation of excavation quantity in open pit mine, and the calculation has high accuracy and stability"

English language is difficult to understand - it is not incorrect, but uncommon words are used and sometimes grammar is not correct.