UAV-Based Multi-Sensor Data Fusion for Urban Land Cover Mapping Using a Deep Convolutional Neural Network
Round 1
Reviewer 1 Report (New Reviewer)
I reviewed the paper in detail and I must say additional experiments and several revisions are required for a possible publication. In current form novelty is limited, data set is small (UAV classical problem), and data augmentation is not mentioned. As the data set is small there is a risk of biased learning. reference methods is too old. My detailed comments are below.
1. Abstract is methodology oriented. It should be providing problem description, the used data and methodology in a much short and concise way and lastly focuse on the findings ad highlight the novelty of research.
2. In the abstract the Authors stated "Unmanned aerial systems (UASs) provide the advantage of flexible and rapid data acquisition at low cost compared to conventional platforms, such as satellite and airborne". These advantages are obvious to some extent, however, the Authors lack to give coverage aspect in this statement. If the study region is large these statements will go in the opposite direction.
3. The main drawback of the paper is that the Authors used a very old "maximum likelihood classification" as a reference for the state-of-the-art algorithms. RS community has not been using this one for a while, and there are plenty of machine learning based algorithms such as SVM or random forest. Authors strongly encouraged to use one of them instead of maximum likelihood.
4. DCNN is given so superficially in the paper. It needs an architectural scheme, model optimization, data augmentation, hyperparameter tuning information should be provided in detail.
5. In contrast data preprocessing is given so long. They should be combined and shortened.
6. maximum classification results are given in whole image based, however DL based method results are given in patches. There is a need for consistency here. Classical method results (one of the methods that I recommend) should be also clipped to patches and we should be seeing comparisons inside the same figures.
7. Accuracy tables should be also combined for classical and Dl based methods for better comparison.
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report (Previous Reviewer 3)
No, further comments
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report (New Reviewer)
Title: UAV-Based Multi-Sensor Data Fusion for Urban Land Cover 2 Mapping Using Deep Convolutional Neural Network
The article is proposed for the special issue "Aerial LiDAR Applications in Urban Environments". The editors suggested that articles will be about the urban environments, especially concentrated since most of the population lives in cities and surroundings. Firstly, the article did not have information about area of interest, where it is, how the urbanization level, etc. The authors should add more information about the urban environmental field or select another journal.
On the technical side:
The abstract is too large. Information from line 17-29 line do not needed. The purpose is explained in a comprehensible manner. The purpose of the article do not explain in the Introduction too. The chapters of the article should be: Introduction; Materials and Methods; Results; Conclusions and Discussion. The authors should correct the structure of the manuscript. Also, the study area is too small or is not explain in the article.
I did not find the discussion chapter and did not understand how the authors compare the got accuracy results.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report (New Reviewer)
Authors performed most of the revisions I requested. I still believe methodology is too long, maybe a flowchart and a shorter version will be better
Author Response
First of all, we would like to thank the reviewer for the careful and insightful review of our manuscript. In the following, we address the comments of the reviewer.
Point 1: Authors performed most of the revisions I requested. I still believe methodology is too long, maybe a flowchart and a shorter version will be better
Response 1: The authors believe that the methodology section is already abbreviated and all the provided information is important for the clarity and completeness of the manuscript.
Finally, we would like to thank the referee again for taking the time to review our manuscript.
Reviewer 3 Report (New Reviewer)
The manuscript is suitable for publication in present form.
Author Response
We would like to thank the referee again for taking the time to review our manuscript.
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
The manuscript is well organized and neatly written. However I regret I could not give a positive decision for its publication in RS.
My negative decision is mainly due to the lack of originality of the present work. In sum, all the related techniques and methods are standard ones, and the authors' chief conclusion that the combination of RGB images with LiDAR could improve the land cover classification seems to me, a foreconclusion. This is because, on the one ahnd, any addition of complementary information will be helpful for the classification, on the other hand, this is amply demonstrated in computer vision community in classification problem, for example, RGB-D outperforms pure RGB input.
6 different combinations are of reference value, but it does not consitute a substantial original contribution per se. I suppose by combining RGB, LiDAR and multispectral images, the result could be further improved.
Anyway, I do not think the manuscript at its present form is publishable in RS
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 2 Report
Recently, research has been conducted to classify various physical environments in urban areas using UAV images and deep learning, and various technologies are being developed to improve accuracy. This study is also considered to be a meaningful study in that respect. The background and purpose of the study are clear, and the methods and procedures accordingly appear to be appropriate. However, since the results of this study can be applied only to the applied target area, it will be necessary to consider the physical environment of various target areas. These aspects should be added and presented as limitations of the study or future plans.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
The author submitted an interesting and well written manuscript dealing with urban land cover mapping. The topic is of great interest for readers as the dataset processed can be used in different applications. However, the authors need to provide some clarifications on different parties of the manuscript. Below are some comments and suggestions:
Line 131: “... six different imaging/LiDAR fusing combinations...“. Please specify if you created these six different images by integrating LiDAR data (not processed using the optimized LOAM SLAM techniques) or point cloud. This would avoid confusion for some readers.
Lines 160-162: Please indicate if during Orthomosaic image generation you created DEM, orthomosaic and DSM as well as.
Lines 173, 179: You cited the Figure 3 before Figure 2. Please check.
Lines 182-184: Please refer to “ Randazzo, G.; Cascio, M.; Fontana, M.; Gregorio, F.; Lanza, S.; Muzirafuti, A. Mapping of Sicilian Pocket Beaches Land Use/Land Cover with Sentinel-2 Imagery: A Case Study of Messina Province. Land 2021, 10, 678. https://doi.org/10.3390/land10070678; Li, Y.; Bai, J.; Zhang, L.; Yang, Z. Mapping and Spatial Variation of Seagrasses in Xincun, Hainan Province, China, Based on Satellite Images. Remote Sens. 2022, 14, 2373. https://doi.org/10.3390/rs14102373“ for recent studies conducted using maximum likelihood in remote sensing applications.
Lines 259-260: Please indicate the number of returns that this LiDAR sensor can acquire.
Line 262: Please indicate the average surface of the area of interest surveyed.
Line 286: Please indicate the band combination adopted for this false colour combination.
Line 289-290: Please indicate if during the point cloud processing some corrections ( trajectories, overlap, noise points,..) were performed to improve the quality to output DEM.
Line 303: Please add the legend indicating the elevation on Figure 11. The Figure looks more likely as DSM
Lines 148, 310: Please specify if you used DEM or DSM.
Line 421: The section of references should be revised.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
I am afraid the authros reply addresses any of my concerns
