Classification of Citrus Leaf Diseases Using Hyperspectral Reflectance and Fluorescence Imaging and Machine Learning Techniques

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Please consider the following issues:

1. In the Sample preparation and measurement, there is a need to add more detail in the identification of the leaves exhibiting the different symptom categories. Are these symptoms identified by expert identification by a Plant Pathologist and Crop physiologist? How accurate is the identification of the citrus diseases and zinc deficiency of the samples?

2. In addition to issue no. 1, there is a need to add a figure in the same subsection containing clear picture examples of the different symptom categories. Figures 4 and 11 cannot clearly see the symptoms of the diseases and zinc deficiency.

3. In Results, please change the sentence construction of the start of paragraphs in the results starting with the word "Figure". For example on page 7,  "Figure 4 shows three types...". "Figure 5 illustrates....". It is monotonous and it lessens the interest of the readers. Please consult a scientific journal writing reference.

4. Scientific names should be italicized in the Introduction. More information is needed to describe the citrus diseases. Indicate the pathogen groups of greasy spot and melanose. 

5. In Results, does the different number of samples for the symptom categories affect the results? 

 

Comments on the Quality of English Language

The quality of English is good.

Author Response

We would like to express our sincere thanks to the reviewers for the critical and constructive comments. These comments are all valuable and helpful for revising and improving our paper. We have studied the comments carefully and made corrections to our manuscript. The following are our responses. All changes in the manuscript and our responses were highlighted in blue.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

 

This work combined Vis-NIR reflectance and fluorescence imaging for citrus leaf diseases classification. A portable reflectance and fluorescence hyperspectral imaging system was used for data acquisition. Nine machine learning classifiers were trained using full spectrum and effective spectral bands. Promising results were obtained, indicating the effectiveness of the integration of the reflectance and fluorescence HSI for disease classification. Overall this work is valuable for biological engineers to develop disease detection systems in an orchard. Some detailed comments are as follows.

1. Figure 5, which sub-figure is the spectral pixel-based spectrum and which one is the leaf-based? Why are there so many spetra in each spectra plot? Please explain.

2. Figure 7, what do you mean by pixel-baed leaf disease? Is it the pixel-level classification (like segmentation)? Regarding the pixel-based and leaf-based, they are very confusing. I suggest to explain them clearly through the manuscript.

3. Section 3.3, the images for prediction in this section are test set?

Author Response

We would like to express our sincere thanks to the reviewers for the critical and constructive comments. These comments are all valuable and helpful for revising and improving our paper. We have studied the comments carefully and made corrections to our manuscript. The following are our responses. All changes in the manuscript and our responses were highlighted in blue.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The paper presents a comprehensive research work. This is very good to build a comparison of many classifiers that are being used in reseach and industry. The final accuracies are not so high but it is expected as there are many inputs and a huge database is needed for a precise classification. I believe the work merits to be published after a minor revision.

Line 40. If available, the production in 2023/24 better to be provided.

 

Materials&methods

Line 136. When abbreviations are introduced, the first letters of words should be capitalized.

Sampling details need to be mentioned (how leaves were chosen, size or age, time to transfer to laboratory etc.) because the phisico-chemical properties of a cut leaf change overtime

Fig. 1. Distance of the camera from sample to be added on the figure or mention in the text

At least basic details of the classification techniques should be added, for example the type of ANN, structure, etc.

Please clearly mention the input/outputs of the classifiers (one disease at a time, or 7 inputs, and the output, healthy/diseased?)

 

Results

Fig. 9. Better to put 0% in the empty squares of the confusion matrix

Results section can be better organized. A table comparing the precisions of all nine methods can be helpful.

 

Discussion

This section needs some work. Dicussion should be extended and comparison with previous research should be added.

Comments on the Quality of English Language

The language of the papper needs an edition. For example line 38, better to change "grappling" with a more common word. Or in line 39, as "etc." has been used, then the "and"before "winter freeze" should be removed.

Author Response

We would like to express our sincere thanks to the reviewers for the critical and constructive comments. These comments are all valuable and helpful for revising and improving our paper. We have studied the comments carefully and made corrections to our manuscript. The following are our responses. All changes in the manuscript and our responses were highlighted in blue.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

Dear Authors,

The manuscript appears to be balanced, well structured and written in readable English.

It discusses the current topic of the development of alternative methods of plant protection against infections. The introduction provides adequate coverage of the topic and justifies the need for the research. The methodology employed in the research is presented in a comprehensive and accurate manner.

 

However, I have some suggestions which can help you to improve the work, please find them below:

 

1. It is proposed that figures 7 and 8 employ bar plots in lieu of continuous point plots. The latter suggests a continuous relationship, which is incongruous with the discrete nature of the data in this instance.

2. Please indicate which input parameters were selected as a result of the PCA analysis.

Author Response

We would like to express our sincere thanks to the reviewers for the critical and constructive comments. These comments are all valuable and helpful for revising and improving our paper. We have studied the comments carefully and made corrections to our manuscript. The following are our responses. All changes in the manuscript and our responses were highlighted in blue.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Line 40 Replace line "Florida is 18.1 million boxes, which is decreased 60% from the 2021-2022 season’s 45.3" to "Florida was 18.1 million boxes, which is a decrease of 60% from the 2021-2022 season’s 45.3".

Lines 470 - 471 Replace sentence "The classification of healthy and disease infected plant leaves has gained attention using machine learning and deep learning methods as reflecting the growth of AI" to "The classification of healthy and infected plant leaves using machine learning and deep learning methods has gained attention reflecting the growth of AI in this field".

Line 483-485 Replace sentences "The deep learning serves complicated high-dimensional data, whereas the machine learning and the application of PCA are straightforward and acquire greater interpretability. Therefore, the machine learning models are not too complicated and effective to interpret the classification of the leaf disease classification." to "The deep learning method is suitable for complicated high-dimensional data, whereas the  machine learning method and the application of PCA are straightforward and have greater interpretability. Therefore, the machine learning models are not too complicated but effective in interpreting the leaf disease classification."

Comments on the Quality of English Language

Minor editing needed especially in the added sentences after review.

Author Response

We would like to express our sincere thanks to the reviewers for the critical and constructive comments. These comments are all valuable and helpful for revising and improving our paper. We have studied the comments carefully and made corrections to our manuscript. The following are our responses. All changes in the manuscript and our responses were highlighted in blue.

Author Response File: Author Response.pdf