PcBD: A Novel Point Cloud Processing Flow for Boundary Detecting and De-Noising

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors propose a novel method (PcBD) for extracting the projection boundary of both rigid and flexible scanning targets that can be used as a process flow of 3D point clouds in 3D target detecting occasions. The network takes the input point cloud and predicts the smoothed projection boundary after Outlier-Removal. The authors introduced a new dataset Bound57 for such tests, each object from the dataset is back projected to simulate real-world scans and randomly sampled at edges to simulate imaging losses. Experimental results show that PcBD framework can distinguish sparse point cloud edges from outlier points in scenarios closer to real scans and is able to predict projection boundaries through 3D information. The proposed framework demonstrated better performance as a process flow for raw point cloud files.

 Comments:

• The authors discussed in Introduction section numerous SOTA methods, where they tried detecting drawbacks of some of them (lines: 54-56, 62-64, 83-85). Finally, the authors presented in lines 86-97 the description of their method that repeats the abstract text. In the opinion of this reviewer, the authors should explicitly highlight the principal contributions of novel framework and explain how it can resolve drawbacks of the existing methods.
• The authors should redact text of their manuscript. This reviewer found that presenting discussion of the SOTA methos in introduction sect., authors again discussed other methods in subsect. 2.2. Feature-Extracting & Outlier Removal (lines 143-150, 155-165). Such a description may be difficult for understanding by a potential reader.
• The authors explained some theoretical basis of their study introducing eqs. 1-11. Please revise these equations and provide the definitions of all the variables used in them. There are a lot of variables that are not defined. For example, there are a lot of the variables that never have been defined in eqs. 4, 6-9. The same comment is also for eqs. 13-15. It would be difficult for a potential reader to understand such equations.
• The authors presented fig.1, fig.2 and fig.3. This reviewer found only short explication for fig.1 in the text of the manuscript, but there did not exist detailed descriptions (or discussions) for other figures. All the procedures presented in Fig. 3. The Feature-Extracting Block should be explicitly explained in the text of this study.
• The authors mentioned in the manuscript the code of the PcBD. This reviewer suggests presenting all the procedures of the proposed framework in the form of the pseudocode summary algorithm that uses equations presented in the theoretical part and can be understandable for a potential reader.
• The authors should use the help of a native speaker to correct several grammar errors, principally in commas.

Comments on the Quality of English Language

The authors should use the help of a native speaker to correct several grammar errors, principally in commas.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

In the article, the authors present the development of the PcBD method for extracting object projection boundaries. This approach utilizes a network that processes input point clouds and predicts a smoothed projection boundary after outlier removal. The method has been tested on numerous real-world examples, demonstrating both theoretical and practical significance.

The article could be further improved by addressing the following suggestions:

Suggestions:

1. The authors should explicitly detail their specific contributions within the introduction section to highlight the novelty and significance of their work.
2. It is advisable to divide the introduction into two distinct sections: one for the general introduction and another dedicated to the literature review. This separation will enhance the clarity and organization of the content.
3. Including a clear and concise problem statement will provide readers with a better understanding of the research objectives and the challenges addressed.
4. Given the extensive use of abbreviations throughout the article, it is recommended to include a list of abbreviations at the beginning for easy reference.
5. The authors should describe the evaluation metrics employed to assess image quality.
6. In the conclusion section, the authors should compare their developed method with existing approaches, providing quantitative parameters to substantiate the advantages and improvements achieved.
7. To strengthen the article's foundation, it is advisable to reduce the number of references and focus on incorporating recent publications from reputable international journals within the last 3–5 years.

Comments on the Quality of English Language

In the article, the authors present the development of the PcBD method for extracting object projection boundaries. This approach utilizes a network that processes input point clouds and predicts a smoothed projection boundary after outlier removal. The method has been tested on numerous real-world examples, demonstrating both theoretical and practical significance.

The article could be further improved by addressing the following suggestions:

Suggestions:

1. The authors should explicitly detail their specific contributions within the introduction section to highlight the novelty and significance of their work.
2. It is advisable to divide the introduction into two distinct sections: one for the general introduction and another dedicated to the literature review. This separation will enhance the clarity and organization of the content.
3. Including a clear and concise problem statement will provide readers with a better understanding of the research objectives and the challenges addressed.
4. Given the extensive use of abbreviations throughout the article, it is recommended to include a list of abbreviations at the beginning for easy reference.
5. The authors should describe the evaluation metrics employed to assess image quality.
6. In the conclusion section, the authors should compare their developed method with existing approaches, providing quantitative parameters to substantiate the advantages and improvements achieved.
7. To strengthen the article's foundation, it is advisable to reduce the number of references and focus on incorporating recent publications from reputable international journals within the last 3–5 years.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

Dear authors,

I have some thoughts and suggestions to offer regarding the paper “PcBD: A Novel Point Cloud Processing Flow for Boundary Detecting and De-Noising” written by Shuyu Sun et al. This study presents excellent results for boundary detection and denoising with point cloud data using by a self-developed PcBD model. Exhaustive research, trend analysis, and detailed theoretical development were utilized by the authors in their systematic description of experimental results. In this regard, the idea and results of the paper can greatly contribute to the development of related fields. Therefore, it is suitable to publish in the present format, except for minor issues. My recommendations for revising the paper are on a separate page.

Best regards.

Reviewer

Additional comments:

 

1. The paper does not have conclusion, so add 5. conclusions

 

2. Figure numbers 13 and 14 are switched, so please correct them and also revise the figure numbers in the main text.

 

3. If you cite a formula, please cite the reference.

 

4. If there are units in the tables, indicate them in units, otherwise express them dimensionless.

Line 11: Please the full name for ToF

 

Line 205: Make sure to put a space after comma

 

Line 234: ...severalother --> several other

Please give definition for the variable N in the root. Could this be 'n'?

 

Line 298: It is not easy to find F_C in Figure 4

• - The End –

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 4 Report

Comments and Suggestions for Authors

The manuscript PcBD: A Novel Point Cloud Processing Flow for Boundary Detecting and De-Noising depicts a research endeavour with an extensive introduction.

However, the text is very imbalanced, having a long description that, one the one hand could be shortened without loss of content but, on the other hand, still misses come critical landmarks and recent research in the field.

Results are shown but analysis is minimal and there are no significant conclusions.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 5 Report

Comments and Suggestions for Authors

This manuscript considers the problem of processing point cloud data from depth sensors, particularly particularly Time-of-Flight (ToF) cameras, where the proposed PcBD method detects boundaries. The proposed PcBD integrates outlier removal, boundary detection, and smoothing into a unified model, addressing key challenges in point cloud quality improvement while preserving the number of points. The authors also introduce Bound57, a benchmark dataset containing realistic synthetic noise, outliers, and labeled projection boundaries, designed to simulate the imperfections typical of ToF camera outputs. The problem considered, it must be stated, is extremely challenging and of timely interest in extracting useful information from wind-tunnel experimentation. In addition, the dual contribution of a multi-task framework and a realistic dataset positions this work as a valuable resource for both research and application in point cloud processing. One of the major strengths of the manuscript is the integrated design of PcBD, which reduces the need for separate preprocessing pipelines and manual parameter tuning. The proposed dataset, Bound57, fills a notable gap in current resources by offering realistic, labeled point clouds suitable for benchmarking both denoising and boundary detection methods. Experimental results demonstrate that PcBD outperforms state-of-the-art methods across various tasks, suggesting that the proposed approach is both effective and versatile. The manuscript is well written and has timely technical contribution. There are, however, a number of comments to which I would like to draw attention of the authors. I was wondering if any comments can be made on the performance of the derived techniques on smoothing inner regions of the point cloud or is it the case that the sole focus of the manuscript is on edges. The projection boundary prediction is not yet generalised across all directions, which may limit its applicability in 3D reconstruction tasks. Are there any comments that can be added in this regard? Despite these issues, the paper offers a clear and novel contribution by combining multiple preprocessing tasks in a single model and releasing a new dataset tailored for realistic sensor conditions. The ideas are practically relevant, especially for robotics and real-time 3D vision applications. To improve the paper, the authors should revise the writing for clarity and grammar, provide more methodological detail (especially regarding the PcBD architecture and its components), and explore the integration of smoothing for inner points and directional generalisation of boundary detection. Evaluation on standard public datasets and the inclusion of ablation studies would further strengthen the empirical claims.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have attended all commnets of this reviewer.