Grape Maturity Estimation for Personalized Agrobot Harvest by Fuzzy Lattice Reasoning (FLR) on an Ontology of Constraints

Round 1

Reviewer 1 Report

This paper introduced a FLR method for grape maturity estimation. Overall, it is an interesting topic and the paper was written in a good flow. The proposed method is well-designed and results are also promising. That said, there are still some minor comments concerning the result evaluation. The sample size of the data to develop the method is still quite small, ~800 for Xinomavro and even less than 100 for the other two cultivars. This consequently jeopardized the validity of the conclusion that the proposed method is better than CNN or any other neural networks as generally, such a small dataset would be insufficient to develop a well-performed deep learning model and thus fail to support a fair comparison of the model performances. Under the circumstance of real-life applications of these methods, considering the normally large size of the data, would the proposed method still outperform a fine-tuned and well-trained CNN? Also, what is the architecture of the CNN model used for comparison? Including more specific details of the compared CNN models would help audience to better understand the strength of this method over other methods.

Author Response

Response to Reviewer 1 Comments

 

Point 1: The sample size of the data to develop the method is still quite small, ~800 for Xinomavro and even less than 100 for the other two cultivars. This consequently jeopardized the validity of the conclusion that the proposed method is better than CNN or any other neural networks as generally, such a small dataset would be insufficient to develop a well-performed deep learning model and thus fail to support a fair comparison of the model performances. Under the circumstance of real-life applications of these methods, considering the normally large size of the data, would the proposed method still outperform a fine-tuned and well-trained CNN?

Response 1: In certain applications, especially real-world applications, only few data are available. The proposed FLRule is a method for dealing effectively with few data, as pointed out in section 3.1 (Discussion). In addition, as Reviewer 3 has pointed out, “… the results aren't the most essential thing; the technique used to obtain these results is”.

 

 

Point 2: Also, what is the architecture of the CNN model used for comparison? Including more specific details of the compared CNN models would help audience to better understand the strength of this method over other methods.

Response 2: Details regarding the CNN architecture have been introduced in the revised manuscript when it is first presented in section 3 (Computational Experiments and Results).

 

Reviewer 2 Report

The manuscript with the title “Grape maturity estimation for personalized agrobot harvest by fuzzy lattice reasoning (FLR) on an ontology of constraints”, proposes and original model for discriminating between ripening grapes in a bunch, that could find applications in efficient and cost-effective machine-harvesting of grapes.  

The model is original, and well-constructed. However, I suggest authors to create (if possible) a flow-chart presenting an outlook from input parameters/criteria to output (discriminative selection). This would encompass the entire general model proposed in a single figure for readers.

The conclusions have 3 paragraphs. However, at the end of the introduction section, the authors presented the structure of the work, rather than aim and objectives. I suggest to express the general aim and then the specific objectives at the end of the introduction. The Objectives are usually defined as the steps proposed for reaching the aim. There objectives should then be mirrored in the conclusions.
Respecting a symmetric structure allows the reader to have a clear red threat to follow when reading the paper from top to finish.

Best regards.

Author Response

Response to Reviewer 2 Comments

 

Point 1: The model is original, and well-constructed. However, I suggest authors to create (if possible) a flow-chart presenting an outlook from input parameters/criteria to output (discriminative selection). This would encompass the entire general model proposed in a single figure for readers. 

Response 1: The requested flow-chart is shown in the new Figure 2 inserted in the revised manuscript with comments shown at the end of section 2.4 (Computational Algorithms).

Point 2: The conclusions have 3 paragraphs. However, at the end of the introduction section, the authors presented the structure of the work, rather than aim and objectives. I suggest to express the general aim and then the specific objectives at the end of the introduction. The Objectives are usually defined as the steps proposed for reaching the aim. There objectives should then be mirrored in the conclusions. 
Respecting a symmetric structure allows the reader to have a clear red threat to follow when reading the paper from top to finish. 

Response 2: The requested aim and objectives are shown in one paragraph in the revised manuscript toward the end of the Introduction section 1 (see the paragraph before the last one).

 

Reviewer 3 Report

Comments on: „ Grape maturity estimation for personalized agrobot harvest by fuzzy lattice reasoning (FLR) on an ontology of constraints”

 Abstract:

The abstract gives an overview of the entire manuscript regarding the motivation of the research, the research methodology and the results supported by the final conclusion.

 Introduction:

The use of agro-robots in viticulture is welcome, considering the paradox of our days: an increasingly large population and, on the other hand, the lack of labour in some fields, implicitly in agriculture.

The ability of robots to selectively harvest grapes is beneficial to producers of special wines who are unable to harvest grapes manually at the optimal time due to a labour shortage.

The arguments for conducting the research on personalised grape harvesting are presented coherently and clearly, which can also be meant by non-specialist readers.

Previous recent researches in the use of modern technology for various works in viticulture, respectively other agricultural sectors, are mentioned in the introduction section.

The objective is clearly defined in the last sentence of the introduction paragraph.

 Materials and methods

 The experimental apparatus is appropriate for the study, especially given that the primary objective of this research is to demonstrate the "power" and utility of the agrobot in personalized grape harvesting according to the level of berry ripening (sugar , acidity, aromatic maturity, etc.).

There is sufficient information provided to replicate the described experiments, but the authors should provide more information about the vineyard location where the samples were collected. Only the region of Eastern Macedonia is mentioned.

Also the authors may wish to mention in this section how the berries samples from those three varieties were collected.  

I don't believe any additional experiments are required to validate the results presented here, because the results aren't the most essential thing; the technique used to obtain these results is.

 Results

The results are clearly presented and in an appropriate format, but there are numerous references throughout the work to previously published results by the same authors (13), making it more difficult to follow the material of the current work.

Why is the chromosome structure for cultivar Xinomavro shown in both a) and b) of figure 2? A brief explanation is required.

These three figures and four tables provide critical information for the material presented and the research objectives.

The data is presented clearly and is not duplicated in figures and tables, which are, after all, simple to understand.

I don't believe any additional graphics are required. However, as previously stated, I believe Figure 2 requires additional explanation.

Appropriate statistical methods have been used to test the significance of the results.

The findings are properly described in the context of the published literature.

 Conclusions

The study's conclusions are supported by appropriate evidence.

The cited literature is well-balanced and relevant to the study.

 

Comments for author File: Comments.pdf

Please refer to the comments in the edited manuscript file for minor comments.

Author Response

Response to Reviewer 3 Comments

 

Point 1: There is sufficient information provided to replicate the described experiments, but the authors should provide more information about the vineyard location where the samples were collected. Only the region of Eastern Macedonia is mentioned.

Also the authors may wish to mention in this section how the berries samples from those three varieties were collected.  

Response 1: The requested additional information about the vineyard locations and the berry samplings has been included in the revised Section 2.2.

 

Point 2: The results are clearly presented and in an appropriate format, but there are numerous references throughout the work to previously published results by the same authors (13), making it more difficult to follow the material of the current work.

Response 2: Redundant references to [13] were removed throughout the revised manuscript for clarity.

 

Point 3: Why is the chromosome structure for cultivar Xinomavro shown in both a) and b) of figure 2? A brief explanation is required.

Response 3: The chromosome structure of Figure 2(b) (i.e. Figure 3(b) in the revised manuscript) is applicable only to the Syrah and Sauvignon Blanc cultivars. The Xinomavro cultivar was shown in the figure caption inadvertently. The word “Xinomavro” has been deleted from the caption of Figure 3(b) in the revised manuscript.