Review of Recent Automated Pothole-Detection Methods

Round 1

Reviewer 1 Report

This manuscript provides a review of established methods for automated pothole detection. It summarizes three different types of detection methods based on the technologies used in pothole detection, including vision-based methods, vibration-based methods, and 3D reconstruction-based methods. Section 2 presents a high-level comparison of the advantages and disadvantages of these three types of methods. In Section 3, recent publications that belong to each type are summarized and are presented in a summary table.

Overall, this manuscript delivers a good summary of recent research in pothole detection. However, the following issues exist.

1. Lack of critical analysis. A review article shouldn’t just summarize the literature. Instead, it must provide in-depth discussions about previous work critically. For example, what is the advantage of this established approach and why it is better than the others? Just saying it is better is not enough, such as Line 166, 206, etc. Moreover, what are the limitations and what can be improved?
2. Lack of background. Even though a brief description of pothole detection methods is presented in the manuscript, e.g., Figure2, it is not sufficient for a review. A review article should provide a reader with basic knowledge in this research area. In this case, for example, what is a typical procedure to perform automated pothole detection? What are the basic scientific principles behind it?
3. Lack of avenues for future research. A review article should identify the research gaps in this area, raise outstanding research questions, and point out potential directions for addressing these problems. A visionary outlook can be beneficial and influential to the further development and expansion of this research field.

Author Response

We appreciate the reviewer’s thoughtful and encouraging comment on our article. We supplemented the paper by reflecting the reviewer’s kind comment. The response to review comment is attached as a file.

Author Response File: Author Response.docx

Reviewer 2 Report

This article falls short of a complete review.  To begin with the topic of how to evaluate the success of pothole detection deserves much more attention.  Instead this topic is glossed over as if it is of little importance.  The manner in which data is summarized also makes it very difficult to compare the various outcomes. The article enumerates the various studies that have been done, but it does not lead to any discussion and conclusion about where the leading edge is currently located and what opportunities present themselves for improvement.   The following are issues that I think need to be addressed:

1. What are the preferred metrics for comparing different pothole detection techniques and is there an agreed upon standard?  Is there a need to have a standard so that various platforms can be evaluated and compared more easily?  Along these lines I would like to draw your attention to: Maxwell, A.E., Warner, T.A., and Guillen, L.A. 2021. Accuracy assessment in convolutional neural network-based deep learning remote sensing studies, Parts 1 and 2 in Remote Sensing. 
2. There is a difference between comparisons relating to accuracy and precision as well as other metrics but this does answer the question as to effective a given process has to be to produce desired and meaningful operational outcomes.  These outcomes include reducing different types of accidents, need for vehicle repairs, and losses in terms of increased travel time.  If the pothole detection rate is 95% is that a meaningful result in terms of its probable impact on operational outcomes?  What is the threshold needed to make this research ready for operational application?
3. By way of background I would also recommend that this topic be placed in the broader perspective of machine learning/deep learning.  A good place to start would be: Zaidi, S.S.A., Anasari, M.S., Aslam, A., Kanwal, N., Asghar, M., and Lee, B. 2022.  A survey of modern deep learning based on object detection models. Digital Signal Processing.  This is important because there may things from the broader perspective that could contribute to making improvements in pothole detection. 
4. There may be other questions that are important in terms of orienting the purpose of this review beyond just enumerating the studies that have been done and doing so in a way that makes it very difficult to make comparisons amongst them.  In this context a good review would I think follow the standard format of background, methods, results, discussion and conclusion.
5. I think you made a good start and that you should not give up on this review.  It is an important topic.     

Comments for author File: Comments.pdf

Author Response

We appreciate the reviewer’s thoughtful and encouraging comment on our article. We supplemented the paper by reflecting the reviewer’s kind comment. The response to review comment is attached as a file.

Author Response File: Author Response.docx

Reviewer 3 Report

The work "Review of Recent Automated Pothole Detection Methods" is descriptive. It concerns the comparison of three automated methods of detecting mechanical damage to the road surface. Automatic pothole detection methods are classified into three types according to the technology used in the pothole identification process. The technologies indicated by the authors are: a vision-based method, a vibration-based method and a method based on 3D reconstruction. In conclusion, the strengths and weaknesses of each method are compared. The potential directions of development of the mentioned technologies in the context of their practical application are also shown.

 

The proposed analysis of the state of knowledge is interesting and fits into the scientific space represented by your journal. However, it requires a few clarifications, additions and possible corrections:

• On what basis were these 3 methods chosen? After all, other modern identification technologies are also used for this purpose ...
• The proposed methods are in fact an application of neural digital image analysis techniques as a classification instrument. Perhaps it is worth highlighting the role of deep learning of multilayer neural networks, which are only mentioned, in the description.
• What is the advantage of the proposed techniques of graphical identification of road damage in relation to other methods?
• The study lacked an unequivocal qualitative assessment of the proposed methods: which is, for example, the best….
• The conclusion was that research on automatic detection of holes should be continued. But it's obvious ……
• Indicate the source of the empirical data used.

Author Response

We appreciate the reviewer’s thoughtful and encouraging comment on our article. We supplemented the paper by reflecting the reviewer’s kind comment. The response to review comment is attached as a file.

Author Response File: Author Response.pdf

Reviewer 4 Report

1. For a review paper, the evalutaion of the method is essential. Only the description is not convining.

2. Some future works should be given in the paper.

Author Response

We appreciate the reviewer’s thoughtful and encouraging comment on our article. We supplemented the paper by reflecting the reviewer’s kind comment. The response to review comment is attached as a file.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The authors addressed most of the reviewers' concerns.

One minor comment:

In lines 191, 210, 243, and 257, there are sentences similar to the following one:

"Authors explained that the performance of a proposed method is better than the performance of existing other methods through the experimental results."

These sentences barely convey any useful information. What are those "existing other methods"? Why the proposed method outperforms them?

Author Response

We appreciate the reviewer’s thoughtful and encouraging comment on our article. We supplemented the paper by reflecting the reviewer’s kind comment. The response to review comment is attached as a file.

Author Response File: Author Response.pdf

Reviewer 2 Report

This is a much improved version. The additional comments concerning strengths and limitations of each study in the text were very helpful.  There are still a few rough spots in terms of English - I have indicated where I think these are and made suggestions as to what to do about it.  I still think more might be done to the tabular summaries - I find them difficult to wade through, but I now think that this should not hold the topic back. 

Comments for author File: Comments.pdf

Author Response

We appreciate the reviewer’s thoughtful and encouraging comment on our article. We supplemented the paper by reflecting the reviewer’s kind comment. The response to review comment is attached as a file.

Author Response File: Author Response.pdf

Reviewer 4 Report

The responses have solved my concerns Therefore, I recommend "accept".

Author Response

We appreciate the reviewer’s thoughtful and encouraging comment on our article. Thanks to the reviewer's thoughtful advice, we were able to review our papers once again.

Author Response File: Author Response.pdf