A Novel Multi-LiDAR-Based Point Cloud Stitching Method Based on a Constrained Particle Filter

Round 1

Reviewer 1 Report

This study proposes a new method for identifying coal piles by LiDAR scanning. The article is suitable for publication in electronics journal in terms of the proposed method and results.

Abstract: Sufficient. It reflects the purpose, findings and conclusions of the study.

Introduction: It can be increased. Although the studies in the literature are discussed in general, the gap in the literature is not well explained and thus the originality of the article cannot be revealed.

Methods: Well explained. The mathematical infrastructure of the method used in the article is well explained.

A discussion section should be added to the article. What the findings from the article mean and their comparison with previous studies should be given in this section.

Conclusion: It should be improved. Too short and not explained well enough.

Author Response

Dear Editor,

Thank you for allowing a resubmission of our manuscript, with an opportunity to address the reviewers’ comments.

We are uploading (a) our point-by-point response to the comments, (b) an updated manuscript with yellow highlighting indicating changes.

Best regards,

Chuanxiang Li et al.

Concern # 1:This study proposes a new method for identifying coal piles by LiDAR scanning. The article is suitable for publication in electronics journal in terms of the proposed method and results.

 Author response: Thank you for the comment.

Concern # 2:Abstract: Sufficient. It reflects the purpose, findings and conclusions of the study.

  Author response: Thank you for the comment.

Concern # 3:Introduction: It can be increased. Although the studies in the literature are discussed in general, the gap in the literature is not well explained and thus the originality of the article cannot be revealed.

Author response: Thank you for the comment.

Author action: We have further elaborated on the innovative aspects and contributions of this paper in the introduction section.

Concern # 4:Methods: Well explained. The mathematical infrastructure of the method used in the article is well explained.

Author response: Thank you for the comment.

Concern # 5:A discussion section should be added to the article. What the findings from the article mean and their comparison with previous studies should be given in this section.

Author response: Thank you for the comment.

Author action: Thank you for your insightful feedback. In response, we have added an introduction section that articulates the significance of our research topic. We have also conducted a comparative analysis of our CPF (Customized Point Fusion) method against previous point cloud stitching techniques, focusing on the comparison of parameter volume and the smoothness of the stitching process. Furthermore, we have discussed the limitations of our current work and outlined plans for future research.

Concern # 6: Conclusion: It should be improved. Too short and not explained well enough.

Author response: Thank you for the comment.

Author action: Thank you for your valuable input. We have revised the conclusion section to better align with the research objectives, contributions, and future work of our paper.

Author Response File: Author Response.pdf

Reviewer 2 Report

In this manuscript, a point cloud generation method is presented. Topic is interesting, but the manuscript should be revised carefully. There are many abbreviations that should be defined in the first position for example LiDAR.

In some parts, radar and in some others LiDAR were mentioned. I did not understand which one is used.

Captions of figures should be presented in a more details for example Figure 1.

Please revise the manuscript carefully from the perspective of grammar.

Colors in figures should be defined in a legend.

Reasons behind selecting methods in the methodology section should be presented.

Validation regarding results is very important. I think the manuscript is not appropriate from this perspective.

Conclusion is not standard. Please mention the aim of study, outputs, and future works here.

Author Response

Dear Editor,

Thank you for allowing a resubmission of our manuscript, with an opportunity to address the reviewers’ comments.

We are uploading (a) our point-by-point response to the comments, (b) an updated manuscript with yellow highlighting indicating changes.

Best regards,

Chuanxiang Li et al.

Concern # 1:In this manuscript, a point cloud generation method is presented. Topic is interesting, but the manuscript should be revised carefully. There are many abbreviations that should be defined in the first position for example LiDAR.

Author response: Thank you for bringing this to our attention.

Author action: We have conducted a thorough review of the acronyms used in the text and have ensured that each one is defined and explained at its first occurrence.

Concern # 2:In some parts, radar and in some others LiDAR were mentioned. I did not understand which one is used.

Author response: Thank you for your observation regarding the terminology used in our manuscript.

Author action: We acknowledge that the work presented in this paper utilizes LiDAR (Light Detection and Ranging), and the abbreviation 'radar' would be inappropriate in this context. We have made the necessary corrections throughout the text to ensure the proper use of 'LiDAR'.

Concern # 3:Captions of figures should be presented in a more details for example Figure 1.

Author response: Thank you for your feedback.

Author action: We have conducted a comprehensive review of the entire document and have added more detailed explanations within the figures to provide a clearer understanding of the content presented.

Concern # 4:Please revise the manuscript carefully from the perspective of grammar.

Author response: Thank you for your careful review.

Author action: We have meticulously examined the manuscript for any potential grammatical errors and have made the appropriate revisions and refinements to enhance the quality of our work.

Concern # 5:Colors in figures should be defined in a legend.

Author response: Thank you for the comment.

Author action: Yes, we have included a representation of the different colored arrows in the legend of the figures to ensure clarity.

Concern # 6:Reasons behind selecting methods in the methodology section should be presented.

Author response: Thank you for your guidance.

Author action: At the beginning of the 'Methods' section, we have critically analyzed the limitations of previous approaches and provided a rationale for our choice of the current method.

Concern # 7:Validation regarding results is very important. I think the manuscript is not appropriate from this perspective.

Author response: Thank you for your insightful comments.

Author action: We acknowledge that the validation of our experimental results was not sufficiently comprehensive. Consequently, we have introduced an additional 'Discussion' section where we have analyzed and compared our point cloud stitching method with previous approaches. This comparison specifically addresses the aspects of parameter volume and the smoothness of the stitching process.

Concern # 8:Conclusion is not standard. Please mention the aim of study, outputs, and future works here.

Author response: Thank you for the comment.

Author action: Yes, we have made revisions in accordance with your suggestions. The Conclusion section has been revised and is now structured to sequentially address the aim of the study, the outputs, and future works.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper deals with an interesting topic of stitching point clouds for coal stockpile monitoring. It is written in perfect English and it is overall well presented.

The introduction gives a good overview of the problem, existing solutions, and their limitations. Radar and LiDAR are somewhat different technologies, but the paper writes about them rather interchangeably. I suggest to make the whole text more consistent, or to explain that the presented algorithms work equally well for both technologies. I appreciate the summary of contributions at the end of Section 1.

The titles of figures 1, 2, 4, 9, 14 start with a small letter. Please fix that. After Equation 9, please describe LLS. I guess it stands for linear least squares.

Section 2.2.1 Voxel filtering, line 177 reads "cubic cells of size 1". What is the unit? How big is the voxel in practice?

Section 2.2.2 Statistical filtering, line 192 reads "nearby points", also Section 2.2.3 Moving least squares, line 203 reads "nearest neighbour set". How do you select those points? Please specify how far apart or how many.

Section 2.2.3 Moving least squares - I understand you take neighbouring points, fit a plane, and project the points on the plane. How does this increase the point density or fill voids? Please describe how do you get new points.

Section 2.3 Edge detection:

Line 217, what is the radius r in practice?

I cannot comprehend v_p in Equation 16. I understand (v_p - p_i) should be a vector starting at p_i and going in the direction of positive z.

I think there should be only two terms in equation 17, one for x and one for y. You can move in x or y and see how z changes.

Equations 18, 19 use N to denote the gradient vector. N is usually used for the normal vector. It's a bit confusing. Also Eq. 19 uses small n. What are the threshold values theta_threshold and epsilon in practice?

Section 3 Experiments:

Line 278 reads "alg. is applied to point cloud 2" but Figure 8 shows cloud 3.

Figure 9 shows the original point clouds stitched together. So the separate filtering described on lines 276-279 was just for demonstration? You first stitch the original point clouds and then apply the filters?

The experiments show good looking results. The quality can be judged mostly visually. There is a table quantifying the smoothness at the stitching seam. You mentioned the voxel filtering is done to reduce the amount of data. It would be nice if you could show how much computational time it spares.

Table 1 has a placeholder in the footer.

Section 3.4 Edge detection is about finding ridges etc. on the surface. But the paper is missing the description how to do that. Section 2.3 only describes the gradient clustering. How is that information used to identify ridges? Section 3.4 only shows the results.

The paper is well written and it solves an interesting topic. The important revision is to describe how to use the clustered gradient information to detect features on the surface.

Author Response

Dear Editor,

Thank you for allowing a resubmission of our manuscript, with an opportunity to address the reviewers’ comments.

We are uploading (a) our point-by-point response to the comments, (b) an updated manuscript with yellow highlighting indicating changes.

Best regards,

Chuanxiang Li et al.

Concern # 1:The paper deals with an interesting topic of stitching point clouds for coal stockpile monitoring. It is written in perfect English and it is overall well presented. The introduction gives a good overview of the problem, existing solutions, and their limitations. Radar and LiDAR are somewhat different technologies, but the paper writes about them rather interchangeably. I suggest to make the whole text more consistent, or to explain that the presented algorithms work equally well for both technologies. I appreciate the summary of contributions at the end of Section 1.

Author response: Thank you for your attention to detail.

Author action:Thank you for your observation regarding the terminology used in our manuscript. We acknowledge that the work presented in this paper utilizes LiDAR (Light Detection and Ranging), and the abbreviation 'radar' would be inappropriate in this context. We have made the necessary corrections throughout the text to ensure the proper use of 'LiDAR'.

Concern #2:The titles of figures 1, 2, 4, 9, 14 start with a small letter. Please fix that. After Equation 9, please describe LLS. I guess it stands for linear least squares.

Author response: Thank you for your attention to detail.

Author action: We have reviewed the captions for the figures in the text and have corrected them to ensure proper capitalization. Additionally, we would like to clarify that 'LLS' refers to 'MLS', which stands for Moving Least Squares method.

Concern # 3:Section 2.2.1 Voxel filtering, line 177 reads "cubic cells of size 1". What is the unit? How big is the voxel in practice?

Author response: Thank you for your observation.

Author action: The size of the voxel has been corrected to 0.3 cubic meters within the text.

Concern # 4:Section 2.2.2 Statistical filtering, line 192 reads "nearby points", also Section 2.2.3 Moving least squares, line 203 reads "nearest neighbour set". How do you select those points? Please specify how far apart or how many.

Author response: Thank you for your feedback.

Author action: For the statistical filtering, we have selected the nearest 100 points as the set of neighboring points. In the case of the Moving Least Squares method, we have chosen points within a radius of 0.5 meters as the set of neighboring points. These modifications have been made in the article.

Concern # 5:Section 2.2.3 Moving least squares - I understand you take neighbouring points, fit a plane, and project the points on the plane. How does this increase the point density or fill voids? Please describe how do you get new points.

Author response: Thank you for your inquiry.

Author action: In the process of upsampling using the Moving Least Squares method, we generate a new point every 5 cm along the xy plane. Given that the coordinates of the xy position and the expression of the fitted plane are known, it is possible to perform interpolation for the void areas.

Concern #6:Section 2.3 Edge detection: Line 217, what is the radius r in practice?

Author response: Thank you for your clarification.

Author action: The actual radius used is 0.5 meters.

Concern # 7:I cannot comprehend v_p in Equation 16. I understand (v_p - p_i) should be a vector starting at p_i and going in the direction of positive z.

Author response: Thank you for your observation.

Author action: The expression (v_p - p_i) actually represents the vector from point p_i to the viewpoint v_i, which we refer to as the view vector. In this paper, the viewpoint v_i is essentially the positive direction of the z-axis. As long as the dot product of the point cloud normal vector and the view vector is greater than zero, which implies that the angle between the normal and the view vector is acute, it ensures that the normals on the coal pile point cloud are all oriented towards the z-axis direction. We apologize for the lack of detailed explanation in the formula and have made the necessary corrections in the manuscript.

Concern # 8:I think there should be only two terms in equation 17, one for x and one for y. You can move in x or y and see how z changes.

Author response: Thank you for bringing this to our attention.

Author action: Yes, we aimed to obtain the gradients of the points in the x and y directions, and we have made the corresponding modifications in the text.

Concern # 9:Equations 18, 19 use N to denote the gradient vector. N is usually used for the normal vector. It's a bit confusing. Also Eq. 19 uses small n. What are the threshold values theta_threshold and epsilon in practice?

Author response: Thank you for your feedback.

Author action: It has come to our attention that using 'N' to denote the gradient vector is indeed not the most appropriate choice. We have replaced 'N' with 'G' to better represent the gradient vector. Additionally, in practice, the threshold values for theta_threshold and epsilon are 15 degrees and 2, respectively.

Concern # 10:Section 3 Experiments: Line 278 reads "alg. is applied to point cloud 2" but Figure 8 shows cloud 3.

Author response: Thank you for bringing this to our attention.

Author action: Yes, Figure 8 was intended to show the effects before and after the application of CPF to cloud3.

Concern # 11:Figure 9 shows the original point clouds stitched together. So the separate filtering described on lines 276-279 was just for demonstration? You first stitch the original point clouds and then apply the filters?

Author response: Thank you for your feedback.

Author action: To eliminate the impact of servo motor rotation errors on point cloud stitching, we first applied CPF filtering to individual point clouds. Subsequently, we performed CPF filtering on the entire point cloud after stitching.

Concern # 12:The experiments show good looking results. The quality can be judged mostly visually. There is a table quantifying the smoothness at the stitching seam. You mentioned the voxel filtering is done to reduce the amount of data. It would be nice if you could show how much computational time it spares.

Author response: Thank you for bringing this to our attention.

Author action: Yes, we have added a 'Discussion' section where we detail the reduction in the number of parameters following voxel filtering. Since the processing time of the algorithm is contingent upon the performance of the device, we have provided the percentage reduction in time, which is a significant decrease of 90%.

Concern # 13:Table 1 has a placeholder in the footer.

Author response: Thank you for bringing this to our attention.

Author action: We have removed the footnote.

Concern # 14:Section 3.4 Edge detection is about finding ridges etc. on the surface. But the paper is missing the description how to do that. Section 2.3 only describes the gradient clustering. How is that information used to identify ridges? Section 3.4 only shows the results.

Author response: Thank you for bringing this to our attention.

Author action: We apologize for any misunderstanding my description may have caused. Our intention was to detect the boundaries of the coal pile to ensure that vehicles can operate within a safe range. Figure 13 illustrates the extracted boundaries, and we have manually delineated the terrain for this purpose. We have revised the manuscript to provide a clearer explanation of this aspect.

Concern # 15:The paper is well written and it solves an interesting topic. The important revision is to describe how to use the clustered gradient information to detect features on the surface.

Author response: Thank you for your feedback.

Author action: We have supplemented and refined the section on the extraction of coal pile boundaries utilizing gradient information.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

I would like to thank the authors for taking into account all the items I suggested in the first revision. It is appropriate for me to publish the article in this form.

Reviewer 2 Report

-