Detection of Road Surface Changes from Multi-Temporal Unmanned Aerial Vehicle Images Using a Convolutional Siamese Network

Round 1

Reviewer 1 Report

Article „Detection of Road Surface Changes from Multi-temporal Unmanned Aerial Vehicle Images using a Convolutional Siamese Network” presents very interesting study regarding monitoring of road condition using unmanned vehicles. And while the article is written very well it could be improved from the pavement side of view. Detailed remarks and questions are given below:

1. Authors described that the method is good in detecting potholes and bigger changes. Also the results concentrate on this. But please describe in details what is the smallest deterioration which can be detected. For example, what is the lowest measured crack width which was detected by the system?
2. Did authors used the method to compare how the method works for more than two images? Like 4-5 consecutive flights?
3. Is system able to detect deterioration in the first flight, or only it is possible in for two photos as a comparison? Is it necessary to make a flight after the construction of the road, or do authors have any proposition for a first reference measurement?
4. Is system able to transfer data into pavement management system? Or give the state in the countable value (like the area of damage, number/length of cracks?)
5. Can you describe in more details, what are the noisy unwanted objects, how weather and road conditions influences the obtained results? What are the recommended weather conditions for the best results in monitoring?

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The manuscript proposes a method for detecting changes in road surfaces using UVA. The idea seems interesting. However, the manuscript lacks in discussion as a research paper and sometimes confusing. Please consider the following comments.

1. Please clearly present its limitation in the introduction. Also, discuss its strength of detecting in terms of quantitative scale.
2. How efficient it is to detect to linear crack (3 mm)? In general the road surface starts with linear crack and later with water penetration the crack leads to potholes and other distress.
3. In general the crack opens more in winter and during summer it contracts. How this method will play role in detecting this changes?
4. How many images were taken for a single crack?
5. How do you select the 163 pairs of small images?
6. Please explain in details the data analysis procedure?
7. Explain the true detection and false detection process.
8. Please explain in details why you are receiving different colors for the detected area and the significance of these color.
9. Is this method comparable to previous study so that the reader have more precise idea about the efficiency of the method?
10. As there is no comparison made, the abstract and conclusion both are very generic.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

The author addressed the reviewer comments.

 

As an introductory study, I appreciate the work. However, the methodology will have significant effect when it can be used for capturing linear crack (3 mm) with seasonal changes.