Simultaneous Localization and Mapping System for Agricultural Yield Estimation Based on Improved VINS-RGBD: A Case Study of a Strawberry Field

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

 

The manuscript is well written but it still needs some minor modifications as follows:

Equations 1, 2, and 3 : describe each parameter

Figures 4, 5, 15 are not clear

Abstract is not matching the title. The abstract is concentrating on best modeling approach for estimating the crop yield, while the title is more generic.

Authors formulated the science gap envisioning their aim to this study by limiting the gap into the lack of positioning accuracy, which is neither not reflected in the abstract nor the title.

 

Comments on the Quality of English Language

Minor editing of English language required

Author Response

Please see the attachment for the reply.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript (agriculture-2961713) proposes a real-time, memory-efficient 3D reconstruction system for strawberry plants using SLAM and an enhanced VINS-RGBD system, paired with the PP-LiteSeg-T semantic segmentation network. This system achieves detailed and accurate semantic point cloud maps, facilitating crop yield estimation and the modernization of agriculture. It significantly reduces memory usage and supports the digital and intelligent transformation of the agricultural sector.

The sections are interesting but verbose. The authors need to focus on the essentials, considerably reducing the manuscript's length to be more concise and direct. All captions must be rewritten as they are unclear, unfocused, and inconclusive.

Consider rewriting the abstract to include introductory information. List keywords in alphabetical order. In the introduction, consider breaking up the text into shorter, more direct paragraphs. Currently, it is verbose and repetitively informative.

The study's objectives should be more direct, focusing strictly on the data actually collected and presented. The authors should describe the captions in greater detail, outlining all elements depicted so that readers do not need to revert to the main text for clarity.

Reconsider the flowchart and explain it adequately. In the materials and methods section, five figures have been presented—why? To make the text more direct and remove unnecessary figures, I suggest relocating them to the supplementary materials.

Figure 6: What does the highlight in red signify? What does the schematic in Figure 8 represent? Why are there eight figures by the results section? Again, in the results section, the authors introduce materials and methods.

The order and labelling of tables and figures need to be revised. Many figures could easily be consolidated. The results and discussion section is inadequate; after a more detailed evaluation, the authors do not discuss the data, leading to excessive description. What were the findings, and how do they support the manuscript’s objectives?

What are the challenges in modelling to create a real-time 3D environment? The conclusions need to be more focused. The bibliography was reviewed, and the most relevant references were included.

Comments on the Quality of English Language

English requires extensive corrections for clarity, grammar, and to remove excessive verbosity.

Author Response

Please see the attachment for the reply.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Real-time Semantic Map Reconstruction System for Strawberry Orchard based on Enhanced VINS-RGBD

The main theme of this research centers on the integration of advanced technologies, notably Simultaneous Localization and Mapping (SLAM), into agricultural practices, particularly in orchard management. The focus is on enhancing efficiency and precision through Semantic SLAM and algorithms, enabling tasks such as monitoring tree growth, estimating yield, and automating processes like harvesting. The research emphasizes the importance of real-time, three-dimensional semantic maps in aiding agricultural robots to understand their environment better, thereby improving task execution. The study highlights the development and evaluation of an enhanced VINS-RGBD system for real-time 3D semantic reconstruction of strawberry plants, which addresses computational challenges, reduces storage needs, and showcases improved segmentation accuracy and real-time performance. Future directions include enhancing segmentation and localization accuracy, potentially integrating the system with UAVs for broader agricultural applications, and aiming to provide real-time, accurate data for agricultural development and economic benefits.

The authors plan to advance this research while focusing on improving the precision of semantic segmentation and localization in challenging surroundings. Additionally, there are plans to integrate the system with UAVs or unmanned vehicles to enable autonomous navigation and map generation. These advancements aim to supply farmers with real-time, precise visual data to aid in agricultural development, facilitating yield estimation and pest monitoring for enhanced economic gains.

Overall, this paper is highly technical basically focusing on the refining of picture-taking with high precision. It lacks a theoretical framework other than image segmentation and minimizing mapping errors. However, it may contribute to the scientific community in imaging agricultural crops, especially orchards.  Hopefully, it will also help in improving food security through precise imaging of agricultural crops at various phases which may help in the application of pesticides and fertilizers. In technical terms, there are no issues other than the English language where sentences are too convoluted in many places requiring serious English language smoothening.

 

 

 

Comments on the Quality of English Language

English needs thorough editing. 

Author Response

Please see the attachment for the reply.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

1. What is the main question addressed by the research?

This work proposes a system that uses Simultaneous Localization and Mapping (SLAM) for the 3D reconstruction of strawberry plants. For this, the authors used an improved VINS-RGBD system combined with the PP-LiteSeg-T semantic segmentation network.  The semantic point cloud map is represented using Voxblox for large scene 3D reconstruction. The authors reached a semantic segmentation accuracy mIoU of 73.2% on a self-constructed strawberry plant dataset, with a processing time of 13.61ms, saving up to 96.91% of memory.

2. What parts do you consider original or relevant for the field? What specific gap in the field does the paper address?

The author's method achieves high position accuracy, which allows them to obtain more accurate maps of strawberry farms. The idea of using a 3D semantic reconstruction system that integrates an improved version of the VINS-129 RGBD system with the PP-LiteSeg-T semantic segmentation network is novel.

3. What does it add to the subject area compared with other published material?

Other authors have presented vSLAM, ORB-SLAM, and DS-SLAM in different scenarios obtaining sundry results. This work used a 3D semantic reconstruction system that integrates an improved version of the VINS-129 RGBD system with the PP-LiteSeg-T semantic segmentation network, which is a novel approach that improves mapping accuracy.

4. What specific improvements should the authors consider regarding the methodology? What further controls should be considered?

In general, the paper demonstrates a strong level of writing and organization, effectively addressing the challenges of the detection problem. However, there are a few recommendations to improve the paper's quality.

- Please, include a dedicated section to explain the dataset distribution and information. This would help in understanding the characteristics of the dataset and how it relates to the proposed methods and results. I recommend using tables to detail the dataset.

- Please, improve the quality of all the figures. For example, Figure 4 is blurred, and the size of the letters is very small compared with the size of the text. Moreover, I strongly recommend using a small figure or image inside each block of the methodology flowchart to improve understanding and to gain better visual representation.

- Please, present metrics of performance for the results of the training of the PP-LiteSeg-T semantic segmentation network. I recommend the use of precision-recall curves for this purpose, which can offer a more comprehensive evaluation of the models' performance.

- I strongly recommend adding a discussion section with all the main findings of the article.

- In the discussion section, please, comment about the possible overfitting problem for the present experiments. 

5. Please describe how the conclusions are or are not consistent with the evidence and arguments presented. Please also indicate if all main questions posed were addressed and by which specific experiments.

The conclusions are consistent with the results presented in this work. The authors reached a semantic segmentation accuracy mIoU of 73.2% on a self-constructed strawberry plant dataset, with a processing time of 13.61ms, saving up to 96.91% of memory. However, a discussion section is still needed to highlight in detail the most relevant findings of this work, as well as future work and improvements.

6. Are the references appropriate?

The references are updated and consider mostly articles from the last 5 years. Moreover, the references match this work topic.

7. Please include any additional comments on the tables and figures and the quality of the data.

It is recommended to vectorize figures 9 and 10 since they have very low resolution and appear very blurry.

 

 

Author Response

Please see the attachment for the reply.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

I appreciate the authors for addressing my queries. The manuscript has significantly improved. However, minor corrections are still needed. For instance, the captions should not be written directly, presenting the sub-items. They need to have a brief explanation of what they are about and then explain what (a), (b)… mean. Moreover, all elements that appear, including colours, should be reported and detailed. The reader needs to have all the information in the captions of the figures and tables, so they don’t need to return to the main text or guess what it is about. Please correct this point in the manuscript. In the material and methods section, the software used needs to be adequately described, identifying the manufacturer, city, state, and country. Pay attention to old bibliographic references and update them appropriately.

Comments on the Quality of English Language

English needs grammar corrections.

Author Response

Please find the attached reply to the reviewer.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The authors have responded to the comments made in the previous review successfully. In particular, improvements have been made in the following points:

- The proposed architecture, along with the details of the database used for this application have been reviewed, and are better explained.

- The quality of the figures, and their explanation has improved significantly.

- Fixed writing errors in the article.

- The conclusions and discussion have been improved, and the article's contribution is explained better.

Based on this, I can say that this paper can be published in its current state.

Author Response

Please find the attached reply to the reviewer.

Author Response File: Author Response.pdf