Effects of Atmospheric Correction and Image Enhancement on Effective Plastic Greenhouse Segments Based on a Semi-Automatic Extraction Method
Round 1
Reviewer 1 Report (Previous Reviewer 2)
All comments have been addressed.
Author Response
Dear Reviewer,
Thank you very much for your approval of our manuscript.
Best regards,
The authors
Reviewer 2 Report (New Reviewer)
The authors try to establish the effects of atmospheric correction and the improvement of the image in effective plastic greenhouse segments based on a semi -automatic extraction method.
The document shows a collection of experiments that seem to be introduced as they were done. In my opinion, the document has several weaknesses:
1. There is no clear methodology that leads to the results.
2. The document is written in a confusing way that hinders understanding.
3. The information provided is not enough to repeat and control the results.
4. The data set is scarce to support the results.
5. Some of the references also seem to have a difficult match in the document.
6. Finally, interest in readers is limited, although the issue covers the scope of the magazine.
In short, in my opinion, the authors present research as a kind of puzzles instead of a clear line of research.
I am sorry to say that the document lacks sufficient solidity to be published as it is.
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 3 Report (New Reviewer)
Dear Authors
Dear Authors
I have now completed the review of the manuscript titled “Effects of Atmospheric Correction and Image Enhancement on Effective Plastic Greenhouse Segments Based on a Semi-automatic Extraction Method”. A new semi-automatic method was also proposed to extract EPGSs in an accurate and efficient way. Firstly, GF-2 images were preprocessed via atmospheric correction, orthographical correction, registration, fusion, linear compression or spatial filtering, and then, boundary-removed point samples with adjustable density were made based on reference polygons by taking advantage of the characteristics of chessboard segmentation. Subsequently, the point samples were used to extract segments containing above or equal to 70% PG pixels in each MRS result quickly and accurately. Finally, the extracted EPGSs were compared and analyzed via intersection over union (IoU), over-segmentation index (OSI), under-segmentation index (USI), error-index of total area (ETA) and composite error index (CEI).
The topic is quite interesting and relevant. I have a few comments to improve the quality and clarity of the manuscript.
- Line 127: Authors orthorectified the GF-2 images with rational polynomial coefficients, please show the RMSE values for the point in a table, like in a supplementary file
- Line 128 “We registered the multispectral image without atmospheric correction (MI) to the corresponding panchromatic image”, which corresponding panchromatic band? pl, add details.
- Line 129: Why authors choose Gram Schmidt Pan Sharpening method over other methods to fuse, please see and add [1,2].
- Table 3. Algorithm parameters and settings in the ESP 2 tool. Is these values fixed for all images captured over different time, please clarify?
- 4.2. Comparison with Related Research, authors have compared only a single study, the comparison should be made with more studies.
- These days ML-based models are used to reap the benefits of computing with better accuracy, authors should add it as a future scope, please also add [3, 4].
- Table 4. In the expression of evaluation parameters, the authors used good metrics for evaluation, however, I suggest that authors should add the computational complexity of the model, see, and add CDLSTM, SMOTEDNN, etc.
- What are the limitations and future scope of the present investigation?
References
[1] How to pan-sharpen images using the Gram-Schmidt pan-sharpen method – a recipe
[2] Fusion-Based Deep Learning Model for Hyperspectral Images Classification
[3] Evaluating the Effectiveness of Machine Learning and Deep Learning Models Combined Time-Series Satellite Data for Multiple Crop Types Classification over a Large-Scale Region
[4] Planetscope Nanosatellites Image Classification Using Machine Learning
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report (New Reviewer)
The Authors have made many changes to improve the paper. In this sense, I just consider minor reasons prior to publishing the paper:
In summary, I suggest authors to change the style of figures. In particular:
Modify Figure 4: avoid the screenshot of the software. Add a table or a new figure of the process.
Figures 5 and 6: Try to shortener the legend. In my opinion the graphical information must be centred. Maybe you can move the legend behind the figure or inside one of the images.
The images contained on figure 8 are different sizes.
Author Response
Dear Reviewer,
Thank you very much for your kind suggestions, which help us a lot to improve the quality of the manuscript.
We have remade Figure 4, 5, 6 according to your suggestion.
As for Figure 8, Figure 8c cannot be shown clearly if its size is minimized than it is now. If we adjusted Figure 8a and Figure 8b to the same size of Figure 8c, the three figures which are related to each other, cannot be shown on the same page. However, we have adjusted the three figures to make their font size seem more similar to each other based on your suggestion.
We hope the reply and the revised version will meet the standards of the journal.
Kind regards,
The authors
Reviewer 3 Report (New Reviewer)
Dear Authors
I have observed that the authors put in good efforts to address all the comments satisfactorily.
Author Response
Dear Reviewer,
Thanks for your suggestion about our English language. We have sent the manuscript to the English professionals for review and made appropriate updates according to their suggestions.
We hope the revised version will meet the standards of the journal.
Kind regards,
The authors
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 idea of evaluating the effects of atmospheric correction and image enhancement on effective plastic greenhouse segments is interesting. However, the paper just simply used the existing methods of preprocessing and segmentation, it obviously lacked novelty to be published in such a highly-influenced journal. In addition, this manuscript had a number of grammar errors and defective sentences, it seemed that it had been submitted without going through a round of proofreading. Therefore, I suggest rejecting the manuscript. I encourage the authors to revise the manuscript with consideration on the novelty and proofreading.
My detailed comments are as follows:
1. Grammar errors:
1) The second sentence (in lines 12-15, page 1) lacks subject, and “make” cannot be listed together with “atmospheric correction”, ……, and “spatial filtering”.
2) The titles of two main parts of technical flowchart (Figure 1) are “Make point samples” and “Image preprocessing” respectively, they cannot be arranged together.
3) The sentence of “Firstly, orthorectify the GF-2 images with rational polynomial coefficients, secondly, register the multispectral image without atmospheric correction (MI) to panchromatic im-124 age, and then fused them by the Gram Schmidt Pan Sharpening tool….” (lines 123-127 in page 4).
4) “Segments overlapping with the reference polygons can be classified by assign algo-…” in line 203, page 6.
5) The sentence in lines 216,217 in page 7.
6) “thus laying the foundation for continuing to use image enhancement to improve the quality of EPGSs.”
7) “thus increasing the computational efficiency of the software” in line 296 in page 11.
….
2. Defective sentences
1) Some long sentences such as “By analyzing…” in lines 280-286 in page 10 and “It can be…” in lines 292-299 in page 11 are difficult for understanding, they need to be shorten.
2) Section 3 abruptly begins with “Firstly”.
3) “Taking the FI as example. “ in line 224, page 7.
4) The sentence of “Meanwhile, the values of USI, ETA and CEI have decreased 0.032, 0.043 and 0.066, respectively, indicating that the atmospheric correction can significantly improve the quality of EPGSs of GF-2 multispectral image, which is consistent with the hypothesis, but the value of USI still has a large room for improvement. “ (in lines 262-265 in page 9) is weird.
…
3. Organization problems
1) The values of gset and parameters of OSP, v1 and etc. (Table 5) should be determined by experiments, thus the last paragraph in subsection 2.4.1 (in lines 194-201, page 6) and the last paragraph in section 2.5 need to be moved to section “3 Results”.
2) The introduction section in general illustrates the study background and purposes, while the authors have listed the experimental results in this section.
4. Figure 8 (c) can be named OSI-USI-ETA-CEI pattern, while Table 4 cannot tell the pattern information, the title of Table 4 should be renamed as “The expression of evaluation parameters (IoU, OSI, … ).
5. “Which” in line 278, page 10 represents what?
6. “Figure 4 shows the specific algorithm steps of a chessboard with 20 pixels as an example:” (lines 219,220 in page 7) should be revised into “Figure 4 shows the specific algorithm steps of making point samples based on reference polygons and …as an example).”
7. The authors need to explain how to obtain the reference polygons.
8. “Finally, the point samples used for extracting PG segments can be derived by export algorithm (Figure 3c)”（line3 210,211, in page 6）, the authors should explain how export algorithm is used to create point samples.
9. “Finally, visual interpretation and manual selection were carried out to extract the EPGSs according to certain rules” (lines 166-167 in page 5), what about the certain rules?
10. The authors mentioned that “the PG segments with g > gset can be defined as EPGSs, the segments with g < gset can be regarded as invalid PG segments“ (lines 174-178 in page 5), how about those segments with g =gset?
11. Figures should be improved:
1) The discarded segment samples are easily confused with the reference polygons in Figure 3.
2) Figure 1.
12. Tables 2 and 3, and Figure 2 are almost the same as those in reference paper [29].
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 2 Report
In this paper authors propose different image enhancement algorithms to improve the multi-resolution segmentation (MRS) quality of plastic greenhouse (PG) in GaoFen-2 (GF-2) images: combination of atmospheric correction, Fast Fourier Transform (FFT), and a circular low-pass (CLP) filter with a radius of 800 pixels. Also, in the segmentation process, "g value" has been set to 70% to have correct final segmentation (small area of extra detected PGs and not detected PGs).
Remarks for the paper:
1. In section 3.2, effect of linear compressing and mean filtering: why did you use a mean filter with kernel size 3x3 (and not for example 5x5 or some other size). Also, what bit depth is your input image (initial bit depth), before you compress the image to 7, 8 or 9 bits?
2. Section 3.3., Gaussian lowpass filter, what was standard deviation and why did you choose its kernel size to be 3x3? Did you try filtering with some other filters in the spatial domain, except mean and Gaussian? For example, a bilateral filter might be also used as a noise-reducing smoothing filter.
3. Section 3.3, probably filter size in the FFT domain might be better explained with its cutoff frequency (i.e. from 0-pi radians in each direction)? Presented filter is probably an ideal lowpass circular filter, did you also try different filters in the frequency domain (for example Butterworth etc.)?
4. Atmospheric correction is also explained in section 4.2, where it is also compared with histogram equalization. Have you considered using adaptive histogram equalization (AHE) or possibly contrast limited AHE (CLAHE)?
5. Overall timing performance should be given in the paper, for all the proposed algorithms, and possibly compared with the initial time needed per tested image.
6. Please better explain "g value", or comment how the segmentation process uses this value? Also, how does the segmentation step segment each image? Have you considered different types of the segmentation that could be possibly used?
Author Response
Please see the attachment.
Author Response File: Author Response.docx
