Tomato Stem and Leaf Segmentation and Phenotype Parameter Extraction Based on Improved Red Billed Blue Magpie Optimization Algorithm

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The research paper "Tomato Stem and Leaf Segmentation and Extraction of Pheno-2 typic Parameters Based on Improved Optimization Algorithm for Red-billed Blue Magpie" is interesting concerning the use of technology and algorithms. However, many issues must be clarified and improved to make it sound more scientific.

1-Rephrase the title instead of using "optimization algorithm for red-billed blue magpie" use "improved red-billed blue magpie optimization algorithm"

2-On line 24 why precision, recall, F1 score, IoU have the same value 0.965?

3-Provide references for paragraphs from lines 41 to 52

4-In lines 87 to 113, the authors should state the problems of Tomato phenological parameters extraction and the hypothesis. Suggesting a new method does not necessarily improve the extraction process. 

They also must specify the deep learning (DL) network with this problem, not all DL networks have the same problem. In the past, some DLs were improved by combining them with other optimization methods.

5-On line 106 use combined or cooperated with deep learning network instead of fusing, and specify the deep learning network name.

6-The 3D data acquisition in Figure 1 shows that tomato plants extracted from the ground to accomplish the task! Is this right or do they remain in the ground and how they were acquired? Were there overlaps in the acquired data? Was this task continued during the plants' growth term?

7-The method does not include details such as:

A-The population size, B-The number of parameters in each individual 

What are these parameters other than learning rate and number of layers?

How the variables of fitness were calculated such as the weights, accuracy...etc are calculated? 

Do you have to run the deep learning for all populations?

No details on 3D CNN such as why the authors selected the simple structure in Figure 2.

8-The authors did not conduct experiments on the speed of the different compared methods.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The paper, "Tomato Stem and Leaf Segmentation and Extraction of Phenotypic Parameters Based on Improved Optimization Algorithm for Red-billed Blue Magpie," addresses segmenting and extracting phenotypic parameters of tomato plants using point cloud data. The authors propose an improved Red-billed Blue Magpie Optimization Algorithm (ES-RBMO) combined with a 3D Convolutional Neural Network (3DCNN). While the topic is relevant and the proposed method is interesting, the paper requires significant improvements.

- The paper is poorly structured and lacks essential elements such as a comprehensive state-of-the-art review, detailed data description, and proper comparisons with existing methods. Additionally, the authors rely solely on virtual plant images rather than real field data, which undermines the practical validity of the study. Below are my main remarks:

 

- The abstract is overly detailed, defining every performance metric unnecessarily. Including only one key metric, such as precision, would suffice.

- The language needs refinement to eliminate grammatical errors and awkward phrasing. For example, "reaches accurate stem and leaf separation without the support of large-scale labeled data" should be rephrased for clarity.

- The introduction lacks a state-of-the-art review of segmentation and classification approaches, especially within the agricultural domain.

- The authors should compare their method with existing methods and highlight its originality.

- This is a data-driven problem, yet the authors provide minimal information about the dataset.

- The paper should include detailed descriptions of the data, supported by real images rather than designed virtual plants.

- The ES-RBMO algorithm is described, but the mathematical formulations (e.g., fitness function weights) lack justification. Why were these weights chosen? A sensitivity analysis is needed.

The 3DCNN architecture is presented without explaining why it was preferred over alternatives like UNet or transformer-based models.

- While performance metrics (precision, recall, etc.) are well-documented, the computational efficiency of the ES-RBMO algorithm is not discussed. How does it compare in runtime and resource usage with baseline models?

- The phenotypic measurements are interesting but require better interpretation. The low R² for stem thickness (0.741) and high RMSE for leaf inclination need explanation.

- The figures are not well-interpreted. For example, Figure 6 requires clearer labeling to distinguish "normal" and "complex" plants.

- Tables lack statistical significance tests, such as confidence intervals, to validate claims of superiority.

- The graphs and images should be better interpreted, and visual comparisons of segmentation outputs should be added.

- The authors should include real field data and compare their method with existing approaches for a more comprehensive evaluation.

- The limitations of the algorithm (e.g., dense growth and occlusion challenges) are understated and need more rigorous analysis.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Dear Authors,

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

The article addresses the current topic of developing accurate methods for image analysis, with a view to supporting vision-based systems for agriculture.

The introduction adequately covers the topic and justifies the need for the research.

 

However, I have some suggestions which can help you to improve the work:

1. Describe the learning set in more detail in section 2.1. Did the set contain 480 images? How was the collection divided into a training set and a test set?

2. Provide the values of the model hyperparameters. Were they adjusted to optimise the model? If so, how?

3. Provide formulas for all metrics used to evaluate the models in Chapter 2.

4. The sentence 'In contrast, although the UNet algorithm[40]...' in lines 228-232 seems to be unfinished.
5. The sentence in lines 236-240 is incomprehensible, it should be rewritten.
6. The results described in sections 3.1 and 3.2 refer to the classification task. In contrast, the results described in section 3.3 refer to the prediction task. Which model was used to obtain these results? Describe this model in Chapter 2.
7. The manuscript does not include Figures 7a and 7b, which the authors refer to in Chapter 4.
8. The manuscript does not include a discussion of the results obtained in comparison to those of other researchers, which is a necessary component of a high-quality scientific paper.

9. Citation should be adapted to the requirements of the journal.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors improved their research paper according to the reviewer's remarks congratulations.

Reviewer 3 Report

Comments and Suggestions for Authors

No suggestions