Pixelwise Complex-Valued Neural Network Based on 1D FFT of Hyperspectral Data to Improve Green Pepper Segmentation in Agriculture

Round 1

Reviewer 1 Report

Review of the manuscript:

Pixelwise complex-valued neural network based on 1-D FFT of hyperspectral data to improve green pepper segmentation in agriculture

1. The findings are sufficiently novel to warrant publication.

2. The conclusions are adequately supported by the data presented.

3. The article is clearly and logically written so that it can be understood by one who is not an expert in the specific field. The work provides an important contribution to its field, consistent with the scope of the journal.

The paper is describing the actual problematics. The aim of the paper is a separating a pepper from densely packed green leaves for automatic picking in agriculture. Given that hyperspectral imaging can be regarded as a kind of wave propagation process. Authors make a novel attempt of introducing a complex neural network tailored for wave-related problems and use Fourier Transform for the construction of complex input. Experimental results have showcased complex neural network outperforms a real-valued one in terms of detection accuracy by 3.9 percent and F1 score by 1.33 percent.

 

Comments:

Discussion is not realized as the comparison with other authors. The statistical methods of the evaluation of data are not described sufficiently. It is suitable to support the results in the Table 1 by some statistical parameters.

Row 50: Please describe the application in the agriculture.

Row 64: Please shortly explain shortcut  SVM.

Row 204: Please describe in detail statistical and quantitative and qualitative differences between methods in the Fig.3

Row 217: Fig.4. Please desribe the axis of the graphs by the quantities and by the units.

Row 233: Please describe comparison and applications in the agriculture

 

 

 

 

 

 

.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The research paper "pixel-wise complex-valued neural network based on 1-D FFT of hyperspectral data to improve green pepper segmentation in agriculture" does not provide new ideas and more serious issues were not handled. Please check the following remarks:

1- The abstract does not indicate clearly the problems encountered in classifying pepper with the currently known methods. Moreover, the authors did not indicate the type of ANN they used and why they have to deploy FFT. Finally, indicate the methods that were compared with the new method.

2-Keywords should be short and representative to the major issues in the paper.

3- The authors raised two important points on lines 31 to 33 by regarding hyperspectral imaging as wave propagation and CVNN that works better with wave-related problems. Please prove that with references

4-Evaluating the literature review one can notice that the combination of CVNN and hyperspectral image to detect pepper is not new and it is mentioned by the authors when they discussed reference [20]. Moreover, the authors claimed that they are the first to use Fourier transform with CVNN to process hyperspectral images. However, checking the literature one can find that CVNN and Fourier transform were used to enhance hyperspectral images.  

Nour Aburaed, Mohammed Q. Alkhatib, Stephen Marshall, Jaime Zabalza, and Hussain Al Ahmad "Complex-valued neural network for hyperspectral single image super resolution", Proc. SPIE 12338, Hyperspectral Imaging and Applications II, 123380H (11 January 2023); https://doi.org/10.1117/12.2645086

5- The datasets consisted of a dataset of pixels instead of sub-images as indicated by the authors because they lack enough images (20 images). The authors have to prove that their idea is feasible using references.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Dear Authors,

thank you for the opportunity to meet the manuscript entitled: "Pixelwise complex-valued neural network based on 1-D FFT of hyperspectral data to improve green pepper segmentation in agriculture".

The subject of the research is undoubtedly interesting and no less important in relation to harvesting a higher crop of green peppers. However, there are several questions regarding the preparation of the manuscript and the applied methodology.

First of all, it is necessary to mention the short duration of the study. Only groundbreaking findings verified over a long time period can be presented in Communication. Despite the fact that experimental results have showcased complex neural network outperforms a real-valued one in terms of detection accuracy by 3.9 percent and F1 score by 1.33 percent, this cannot be considered groundbreaking.

Moreover, a big question is the sample used to evaluate the performance of the monitored method. In my opinion, this is a serious flaw.

There is also room for improvement in the references. Despite the fact that this is a short communication, it is advisable to give room for discussion and comparison of own results with existing ones.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

The authors replied to every point raised by the reviewers. Although some replies do not sound scientific, I consider the overall modification sufficient for accepting the paper. I should point out that the objectives of the reviewer are not only to judge the paper but also to improve the authors' work. Congratulations on your research work.

Reviewer 3 Report

The authors sufficiently responded to my comments and modified the manuscript based on some of them. However, based on my initial review, I insist that this manuscript cannot be considered a Communication, but a Brief report could be more appropriate.