An Integrated Method for River Water Level Recognition from Surveillance Images Using Convolution Neural Networks

Round 1

Reviewer 1 Report

The authors present an image detection method for automatic reading water level gauges. The results shown that the model is working well.

I have no further comments on this manuscript.

Author Response

Thanks for your positive comments.

Reviewer 2 Report

Hello to dear authors

Thank you for your efforts to improve neural networks and artificial intelligence. I think your research is very useful and interesting. There are some serious structural and grammatical corrections to your manuscript. I also asked a few questions about the methodology and validity of the results. Please read the comments and correct them carefully.

(1) Based on my impression you need to revise the manuscript's title. my suggestion : Development of an integrated method for river water level reading from surveillance images using convolutional neural networks (CNN).

(2) Section #2 was poorly presented. Please provide a flowchart or schematic overview of all the equipment and procedures you described. This section is already very confusing for readers.

(3) The number of references should be at the end of the sentences, not in the middle.

(4) Overall, the level of English in this manuscript is unacceptable (as an academic presentation). You should have it completely revised by a native English editor. Below I have pointed out some English and structural mistakes:

Line 7 : "Firstly, we improve the7 FCOS mode..., it should be => The FCOS mode was improved....

Line 12 : "we combine all the results..." , it should be => All of the results were combined... .

Lines 13 & 14 : "the problem of water gauge water level...' what is the meaning of water gauge water level ? please rewrite this sentence correctly.

Line 16: same to Line 12 error.

Line 17: you should not use same words in the title and the keywords.

Line 20: you need more references here.

Line 21 & 22 : The floods occurred just in rainy seasons? please give some references, or revise the sentence.

Line 22: According to the reports not statistics.

Lines 22, 23 and 24 : you should give references about this reports exactly.

Line 25 : please give references here.

Line 26 : Level detection or recognition? please use one word to talk about the aim throughout the manuscript.

Line 27 : Other electronic equipment(s).

Line 27 : you need some references.

Line 30 : The level of the water can be recognized by ...

Line 31 : Which method has advantages and disadvantages?

Line 36 : You need some references.

Lines 36 to 38 : You need some references.

Line 39 : You need some references.

Line 52 : vague sentence : The first is to locate the area with water gauge from image.

Line 60 : you should not use "we" or "i" or "our" and etc. please check this note all over the text carefully.

Line 55 : Please check the sentence grammatically.

Line 53 : what was your mean to use these words : "easer image" and "eage detection"?

Line 57 : please put the reference number at the end of the sentence.

Line 58 : You need some references.

Lines 59 & 60 : Confusing sentences : Usually, the sliding window method is used to generate an imprecise region as a candidate region, then HOG is used to extract the features we need of candidates, and finally SVM is used to judge whether the region contains water gauge.

Line 72 : you need some references about your claim.

Line 82 : incomprehensible and unnecessary sentence.

Line 83 : the wrong way to give reference's numbers. [10][11][12] is incorrect, please correct them to this form : [11-12]

Line 84 : other fields such as mineral exploration and geological studies. two references about this claim which you can use them:

1. Camps-Valls, G., Tuia, D., Zhu, X. X., & Reichstein, M. (Eds.). (2021). Deep learning for the Earth Sciences: A comprehensive approach to remote sensing, climate science and geosciences. John Wiley & Sons.‏

2. Shirazy, A., Shirazi, A., & Nazerian, H. (2021). Application of Remote Sensing in Earth Sciences–A Review. International Journal of Science and Engineering Applications, 10(5), 45-51.
3. Chen, Q., Zhao, Z., Zhou, J., Zeng, M., Xia, J., Sun, T., & Zhao, X. (2021). New insights into the Pulang porphyry copper deposit in southwest China: Indication of alteration minerals detected using ASTER and WorldView-3 data. Remote Sensing, 13(14), 2798.‏

Line 87 : you need some references.

Line 93 : you need some references.

Line 94 : please put the reference's numbers to end of the sentence.

Line 96 : who claimed first time? please give a cite...

Lines 98 & 99 : "Conventional linear-type detection98
methods cannot solve this problem", what problem? how to find the water level line? is it a problem or a key step? and finally please tell me why it can not solve the problem? and give some references here.

Lines 104 and 105 : based on which study by who?

Line 121 : you used "we" and "our" again. as i said before, please check all over the text carefully.

Lines 128 & 129 : "The water gauge above or under the water in the water gauge image have the same image features" it is a confusing sentence grammatically. please rewrite it.

Line 130 : what is it? "underwater water".

Line 162: "point's" ?

Line 236 : Inappropriate title selection, you need to revise it academically.

Line 256 : please mention the full name of the method : Fully Convolutional One-Stage (FCOS). Pay attention to this point at the beginning of the manuscript.

Line 258: Do the same as in the previous comment for FCN.

Figure 1: Where did you write about the car in the text? Why did you write Chinese in this form? Please make a schematic about your work, not about other things.

Lines 276 to 279 : The description of your improvement is unclear and insufficient. Please write this part more carefully and explain exactly what your creativity was in the methodology.

Figure 4 : please write a comprehensive explanation about the process that applied at this stage in the caption.

Lines 305 and 306 : "The water level line extraction schematic is shown in the Fig.4." => The water level line was shown schematically in the figure 4.

Line 235 : "Method" section => Raw Data & Methodology (or Technical Flow). you should prepare a flowchart and show all steps from the start to the end. and highlight your innovation in the flowchart. finally put it at the end of methodology section.

Line 332 : please move the sub-section 4.2. Dataset to the Raw Data & Methodology section.

Line 337 : you have to cite about internet source. please reference exactly to the websites or reports.

Line 415: Here you said about manual reading. Please explain clearly about these readings. As I understand it, you used the comparison between manual recognition and algorithm recognition to confirm and evaluate your final results. So it is very important to understand exactly what manual detection means?

Line 434: please write a sentence about your method's shortcomings briefly.

Line 427: You need a section titled "Discussion" before the "Conclusion" section. In this section you should discuss the results and a brief comparison with other related research.

As a personal question : What kind of project was the result of the research presented in this manuscript? Academic or industrial?

Best Regards

Author Response

Please check the attached file for review.

Author Response File: Author Response.docx

Reviewer 3 Report

This manuscript adopted convolutional neural networks for river level recognition based on surveillance images, where deeplabV3 framework was established for the task of interest. The experiments were conducted to validate the performance of deep learning models via a comparsion with other segmentation networks, with satisfactory results. Overall, the topic of this research is interesting, and the manuscript was well organised and written. The detailed comments are summarised as follows.

1.       The contribution and innovation of the manuscript should be clarified clearly in abstract and introduction.

2.       Broaden and update literature review on CNN or deep learning in engineering applications. E.g. Torsional capacity evaluation of RC beams using an improved bird swarm algorithm optimised 2D convolutional neural network; Vision-based concrete crack detection using a hybrid framework considering noise effect.

3.       The performance of deep learning models is heavily dependent on the setting of hyperparameters. How did the authors set them in this research to achieve optimal segmentation performance?

4.       The training process and results should be presented, such as accuracy and loss.

5.       How about the robustness of the proposed method against noise effect?

 

6.       More future research should be included in conclusion part. 

Author Response

Please check the attached file for review.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Hello dear authors

thank you for detailed answers on the review report.

Please consider these notes :

(1) please replace Figure 1 in the last version of manuscript (Currently not replaced).

(2) The English language should be polished carefully. Unfortunately, the English language is currently unacceptable.

Best regards

Author Response

please check the attached file for review.

Author Response File: Author Response.docx