Texture Analysis to Enhance Drone-Based Multi-Modal Inspection of Structures

Round 1

Reviewer 1 Report

Dear authors,

First of all, I would like to congratulate you on the work you have done. Multi-sensor data fusion is a novel field that helps organisations to work optimally with current and future tools.

The use of platforms such as UAVs allows cost savings in infrastructure maintenance and their work can help to make it even more efficient.

Allow me to ask you, in my humble opinion, to resolve a number of doubts and to make a number of comments on your work.

I will be very interested for further clarification on what a thermal camera is, and what information can be obtained from this type of sensor.

 Are there previous works and specific references on the fusion of thermal and visible images? Please describe if any previous attempts have been made to fuse these types of images.

In lines 136-137, you affirm that the geometrical alignment between the thermal and visible images is a critical point in the pre-processing phase. What are the problems of misalignment in this phase? I would like to know more about how the alignment of images is performed.

 

In the enumeration of image fusion techniques (lines 140-142), you list several different methods. According to those listed, many of them appear to be pansharpening algorithms. Could you elaborate more on this aspect and how these methods work?

 

What do you mean by "light spectrum"? line 159

 

Please justify the reason why thermal images enhance the defects in the absence of illumination. Is this true in all cases? In all situations? In all materials? lines 273-274

 

In this paragraph (lines 291-301) it is not described what the product of this phase is. Is it a multispectral image? Please clarify this.

 

About the dataset mentioned (Line 307), what is it composed of? What kind of images does it contain? What kind of sensors? Are the images contained in this dataset already aligned? 

 

It is not clear from Figure 4 what the last column on the right represents.

 

What do you mean by "manual registration"? line 341. How is this process carried out? What is needed to perform this phase? What is the final product of this procedure? Please describe thoroughly how it is done so that readers can follow the procedure.

 

The same as before with the concept “manual alignment”. Line 386

 

Please describe what this dataset contains (lines 394-402).

 

I feel insulted that a table in a candidate paper for a scientific publication contains units in the imperial system of measurement (miles, inches, Fahrenheit, etc). Really? The use of these imperial units should be eradicated from any study or work in science and engineering. Please express the units in these tables 4 and 5 in the International System of Measurement (SI), which is the only accepted by the scientific community.

 

Supplementary material: it is not possible to access the code, dataset, etc. because the links are broken. Please check that this supplementary material is accessible.

 

As a final remark, I consider that the structure of the written work is not correct. It is very difficult to follow the different cases and the results obtained. I suggest a total editing of the structure of the article to allow comparison and analysis of the results obtained. Please include more pictures and examples of fault identification obtained for comparison. I also suggest you express more clearly the quantitative analysis of the results obtained. It is not clear whether and to what extent the proposed models are adequate. A table expressing the affirmative cases of identification with respect to the whole set would be very desirable.

 

 

 

Author Response

We would like to express our sincere gratitude to respected reviewer for taking time from his/her busy schedule to review our paper and we are grateful for his/her comments. As suggested by the reviewer, we have carefully revised the paper based on your comment. The complete response letter to the reviewer's comment is attached in the following.

Moreover, the changes are highlighted in the revised manuscript and the list of changes are embedded in the paper (under abstract section).

Author Response File: Author Response.pdf

Reviewer 2 Report

I suggest rewriting the abstract and making it going straight to what is necessary along with adding the main numerical results.

In line 90, I suggest removing the expression "such as drones".

In line 212, Why have you chosen ResNet-50 as a UNet encoder? The small size is not a viable argument as there is a lighter version (ResNet-18), as well as other lighter encoders such as EfficientNet-B0, DenseNet-121. The proven feature extraction capabilities lack justification.

In line 282, it is stated the visible images are pre-processed to balance illumination and contrast. Why such operation has been performed? and what are the parameters adopted for this preprocessing?

Generally, thermal infrared and RGB cameras have different focal lengths and lens distortions as well as different resolutions. How the authors suggest addressing this issue?

The text in diagram figures has a very small size and is difficult to read.

Author Response

We would like to express our sincere gratitude to respected Reviewer for taking time from his/her busy schedule to review our paper and we are grateful for his/her comments. As suggested by the reviewer, we have carefully revised the paper based on your comment. The complete response letter is attached in the following.

Moreover, the changes are highlighted in the revised manuscript and the list of changes are embedded in the paper (under abstract section). 

Author Response File: Author Response.pdf

Reviewer 3 Report

Dear editors,

 

The manuscript titled “Texture Analysis to Enhance Drone-based Multi-Modal Inspection of Structures” proposed process pipeline textually segments the visible images and uses the result to improve the detection and characterization of possible abnormalities in thermal images. Moreover, authors introduced four case studies with different process pipelines and datasets to demonstrate the benefits of the presented approach in various industrial applications. The contents are rich and wonderful. However, whether the data process pipeline or the methods used in the four case-study are original, that is to say, the algorithms are state-of-art methods. Therefore, the originality of the ‘article’ is questionable. According to the content, I would rather consider the manuscript ‘a survey’ or ‘review paper’. Moreover, some seemingly insignificant unregulated tables, figures and web-links but reflecting the quality of the presentation are listed below:

1.       The format of Table 2-5 does not conform to the template.

2.       The font size of the texts in the flow charts, Fig.3 and Fig.5 could be bigger so that readers can read them easily.

3.       In figure 14, the styles of the three subplots are not consistent with each other. 

4.       The two web-links below line 582 could not access normally.I find the response web-pages are 404 labeled.  

 

Finally, references 9 and 10 are self-citations listed below.

Ref.9.  Nooralishahi, P.; Ibarra-Castanedo, C.; Deane, S.; López, F.; Pant, S.; Genest, M.; Avdelidis, N.P.; Maldague, X.P. Drone-Based Non-Destructive Inspection of Industrial Sites: A Review and Case Studies. Drones 2021, 5, 106.

Ref.10. Alhammad, M.; Avdelidis, N.P.; Deane, S.; Ibarra-Castanedo, C.; Pant, S.; Nooralishahi, P.; Ahmadi, M.; Genest, M.; Zolotas, A.; Zanotti-Fragonara, L.; et al. Diagnosis of composite materials in aircraft applications: towards a UAV-based active thermography inspection approach. In Proceedings of the Thermosense: Thermal Infrared Applications XLIII. SPIE, 2021, Vol. 11743, pp. 35–41.

 

Would the authors please explain the essential differences of these ‘pipelines? Maybe the methodological innovation could be appropriate to be addressed in detail.   

--Comments Over--

 

Author Response

We would like to express our sincere gratitude to respected Reviewer #3 for taking time from his/her busy schedule to review our paper and we are grateful for her/his comments. As suggested by the reviewer, we have carefully revised the paper based on your comment. The complete response letter is attached in the following.

Moreover, the changes are highlighted in the revised manuscript and the list of changes are embedded in the paper (under abstract section).

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Dear author,

Anyway, I still do not think the work in the article relates to innovation. The hardware innovation is from the DJI M300 drone equipped with multimode imaging sensors. Except for the multi-modal system with the intrinsic capability of multi-modal inspection, the demonstration of the four cases certainly justified the function effectivity of the DJI drone. However, is it could be considered an innovation of the work in this manuscript? Therefore, it is hard for me to assure that the manuscript is a research paper.

Of course, the contents of this manuscript are enough to be a technical report. I acknowledge to some degree that the authors argued in the response letter that it is unfair to consider this work to be a review or a survey. If calling it a technical report, it may be more moderate and acceptable to some readers.

The authors have revised the paper according to some of my suggestions and a lot of work has been done for the manuscript. Therefore, a positive decision could also be moderate. 

Author Response

We would like to thank respected Reviewer #3 again for taking time from their busy schedule to review our revised paper. We also want to express our gratitude for their comments. In order to fully address the reviewer's comments and provide a more clear explanation of the paper's originality. The paper is revised accordingly. The changes are highlighted in the revised  manuscript and the list of changes are embedded in the paper (under abstract section).

Moreover, a response letter is prepared to answer the comments further and attached accrodingly.

Author Response File: Author Response.pdf