Luotuo Mountain Waste Dump Cover Interpretation Combining Deep Learning and VDVI Based on Data from an Unmanned Aerial Vehicle (UAV)

Round 1

Reviewer 1 Report

Dear authors,

Your manuscript is on a very interesting topic. I have these follow up comments:

1. I would recommend adding the full names of the abbreviations appeared in your work, such as “SR” “BP” “GAN”, as they can be blurring for beginners of this field.

2. Section 3.1:  “… Section 3.1.1 presents the Mobilenetv2 backbone feature extraction network, Section3.1.2 briefly covers the atrous convolution method, and Section3.1.3 provides a detailed description of the structure of our model.” I believe the author made a mistake in the description of chapter arrangement.

3. Fig.12 and Fig.13: A legend is required to illustrate the segmentation result, which color stands for vegetation, roads, and water respectively?

4. Fig.12: (a), (b) and (c), (d) represent the “two typical mining gangue hills” respectively? If so, as I can see the site shape in (c) and (d) is identical to the study site illustrated in Fig.1, is the other site also shown in Fig.1? If not, I would recommend illustrating both mining gangue hill sites. Please clarify.

5. Table 2 only listed the information of three algorithms, I don’t think the name of table 2 is appropriate.

6. Section 4.1: (a)“We stitched the flight data taken by the UAV at 192m altitude and cropped a total of 1350 images of 1500*1000 size for training, used the dataset of 384m flight for hyper-segmentation experiments”, the sentence here would cause confusion at first. I would recommend moving this sentence to 2.2 where a more detailed illustration should be given on how each part of the data set is applied in SR reconstruction and ground segmentation. (b) Camelot dataset were in the same time scale with dataset of 384m flight? Its role should be more detailed.

7. Section 4.1: (a) REAL-SR was compared with three methods which are ESRGAN, SRMD and WAIFU2X, why were the three methods chosen for comparision? Their common basis of comparison should be explained. (b) “MORGAN” in table1, a mistake? Should it be ESRGAN?

8. Section 4.2.6: The description of the results should be based on table 6, not table 4. Please make a carefully check throughout the paper to ensure the right correspondence between text and charts.

9. Section 4.3: (a) “As can be seen from Table 6, among the visible vegetation indices…”, I think it’s table 7 that should be refered to here. (b) The conduction of SVM and K-means methods based on what images? Please give more detailed information.

10. Conclusions: (a) As far as I can see, HR images using SR reconstruction were not been applied in ground segmentation? Cause section 4.1 mentioned: “used the dataset of 384m flight for hyper-segmentation experiments”. I think the relationship between SR reconstruction and ground segmentation should be further discussed to make a more complete and coherent logic. (b) “However, all in all, using super-resolution reconstruction in remote effectively reduces the aircraft failure rate and flight distance, thus significantly improving efficiency.” I can’t follow. Why does the SR reconstruction can reduce the aircraft failure rate? By reducing operation frequency? Please revise the sentence and get to the point appropriately. (c) Please focus on what the audiences can learn from this study. Please reduce the sentences indicating what has been done.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

OVERALL: - Please, enlarge and re-arrange all Figures and font sizes in order to guide the reader properly in all sections. All figures must be composed of HD images. It is mandatory to improve the scientific quality of the whole manuscript.

-        Please, pay attention to the JOURNAL TEMPLATE in all sections.

INTRODUCTION: Please, consider in the scientific background of your work the most advanced remote sensing methods employed in the management and monitoring of vegetated water resources (i.e.,

Lama, G.F.C., Errico, A., Pasquino, V., Mirzaei, S., Preti, F., Chirico, G.B. 2021. Velocity uncertainty quantification based on Riparian vegetation indices in open channels colonized by Phragmites australis. J. Ecohydraulics 7(1), 71–76. https://doi.org/10.1080/24705357.2021.1938255.

Jalonen, J., Järvelä, J., Virtanen, J. P., Vaaja, M., Kurkela, M., & Hyyppä, H. (2015). Determining characteristic vegetation areas by terrestrial laser scanning for floodplain flow modeling. Water, 7(2), 420-437. https://doi.org/10.3390/w7020420.

Yaseen, Z. M. (2021). An insight into machine learning models era in simulating soil, water bodies and adsorption heavy metals: Review, challenges and solutions. Chemosphere, 277, 130126.

Khan, M.A., Sharma, N., Lama, G.F.C., Hasan, M., Garg, R., Busico, G., & Alharbi, R.S. (2022) Three-Dimensional Hole Size (3DHS) Approach for Water Flow Turbulence Analysis over Emerging Sand Bars: Flume-Scale Experiments. Water, 14, 1889. https://doi.org/10.3390/w14121889)

 

METHODS: Please, re-arrange all Figures in each sub-section. They are too poor at the moment; there is a loack of information.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Minor editorial errors are highlighted in the attached file.

Paper may be accepted for publication after minor revision .

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 4 Report

Segmentation of remotely-sensed images is a huge and well-studied area. An enormous number of techniques have been developed, some of them offering very good classification accuracy. But different application areas require their own image classification algorithms - what's good for one application is not necessarily good for another. This paper describes a particular problem of ground vegetation cover classification in a fairly inhospitable environment where it is important to determine the extent of re-vegetation on gangue hills. Recognising grass cover in this environment is not easy.

So, the problem being tackled here is an important one. And the images being processed for this study are not from multispectral sensors on satellites - instead they are generated by UAVs. The approach taken in this research seems very sensible and useful results are presented in this paper. The authors also recognise that there is also much work left to be done.

This is a well-structured and well-written paper. The paper starts with a clear introduction explaining the problem and justifying the approach taken, particularly showing why a UAV had to be used. There is a useful description of related work in the field. The methodology is described well with a clear description of the processing steps required. The results are presented clearly and the claims made by the paper appear to be justified in the light of the results. There are sensible conclusions and a good list of references. The paper is well-illustrated.

I have just a few, essentially typographical, suggestions.

Make sure that there are spaces after sentence-ending full stops, throughout.

Title: from Unmanned Aerial Vehicle -> from an Unmanned Aerial Vehicle

p1 - 1.1 gangue hill for -> 1.1 Gangue hills and

p1 - Shanxi is mentioned twice!

p2 - in disadvantage 2 - in the gangue mountain used for landforms change -> if the gangue mountain used for landforms changes

p3 - of satellites in orbit -> in images from satellites in orbit

p4 - Astragalus. Agriophyllum -> Astragalus, Agriophyllum

p6 - The data was then splited -> The data was then split

p9 - retains the most range of feature -> retains the greatest range of feature

p10 - we convolved the X wth two -> we convolved the X with two

p10 - 4 result -> 4 Results

p13 - We gave a selection of -> We show a selection of

p14 - because Mask-RCNN segments -> because Mask-RCNN attempts to segment

p15 - The information in table 2 can just be listed in the text. Similarly for table 4. And the two mentions of "Ubuntu 18.04 operating system and an NVIDIA RTX 5000 graphics card" can be removed because it has already been stated.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

The article has been improved in the current version. There are still few issues to be fixed.

 

INTRODUCTION: Please, consider in the scientific background of your work the most recent improvements of machine leaning in water resources applications broadly speaking (i.e.,

 

Lama, G.F.C., Errico, A., Pasquino, V., Mirzaei, S., Preti, F., Chirico, G.B. 2021. Velocity uncertainty quantification based on Riparian vegetation indices in open channels colonized by Phragmites australis. J. Ecohydraulics 7(1), 71–76. https://doi.org/10.1080/24705357.2021.1938255.

 

 

Kutz, J. N. (2017). Deep learning in fluid dynamics. Journal of Fluid Mechanics, 814, 1-4.

 

 

Shen, C. (2018). A transdisciplinary review of deep learning research and its relevance for water resources scientists. Water Resources Research, 54(11), 8558-8593)

 

After these corrections the article is suitable for publication.

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx