High-Throughput Legume Seed Phenotyping Using a Handheld 3D Laser Scanner

Round 1

Reviewer 1 Report

This work proposed a pipeline of automatic data acquisition and processing, including point cloud acquisition, single seed extraction, pose normalization, 3D reconstruction and trait estimation.  Although the overall average reconstruction error is low, the authors are highly required to address the following issues:

Questions & Mistakes:

1. Figure Mistake: (page 4) figure 2 is not clear and the lines overlap with the
2. Figure Mistake: (page 6) figure 4 lacks Z-axis.
3. Typographical Mistake: (page 7) line 238 and line 239 are blanks.
4. Question: Why only acquire five common dry legume seeds?
5. Question: How will results change if legume seeds aren’t dry or fresh?
6. Question: Why does the obtained point cloud have no color information since using a SmartSCAN3D-5.0M color 3D scanner together with an S-030 camera?
7. Question: Does the angle of a handheld 3D laser scanner heavily influence the results? Is it possible to change to a fixed angle scanner?

Author Response

Please see the attachment because of there are graphics.

Author Response File: Author Response.docx

Reviewer 2 Report

The paper presents a new method for high-throughput legume seed phenotyping. The seeds of soybeans, peas, black beans, red beans, and mung beans were applied. However, only 100 samples of each type of seeds were used. In my opinion, the number of samples is not enough. A novel method should be tested and validated using more samples including more external sets. Additional experiments are needed. The research material is not sufficiently described (origin, varieties, growing conditions). The presented seed images are of low resolution. I am not sure whether the segmentation was done correctly. Depending on the species of seed, the background of the image could be changed. The results are not sufficiently discussed against the background of the available literature. The Conclusions section provides the same information as the Abstract.

Author Response

Please see the attachment because of there are graphics.

Author Response File: Author Response.docx

Reviewer 3 Report

The topic of the manuscript is of interest and the proposed method is innovative. The abstract is very well written and organized, however the hypothesis is not clear from the abstract. The English language is also fine. Congrats to the authors. Below are some minor points to improve your manuscript.

 

• Please add one or two lines to the end of the abstract stating how the proposed method can be used for seed phenotyping.
• Please add one or two lines to your literature review (lines54-71) about topics related to structure from motion and the impact of camera viewing angles for estimating agricultural parameters from 3D point clouds. You can cite the following references:

Li, M., Shamshiri, R. R., Schirrmann, M., & Weltzien, C. (2021). Impact of Camera Viewing Angle for Estimating Leaf Parameters of Wheat Plants from 3D Point Clouds. Agriculture, 11(6), 563.

Harwin, S., Lucieer, A., & Osborn, J. (2015). The impact of the calibration method on the accuracy of point clouds derived using unmanned aerial vehicle multi-view stereopsis. Remote Sensing, 7(9), 11933-11953.

De Souza, C. H. W., Lamparelli, R. A. C., Rocha, J. V., & Magalhães, P. S. G. (2017). Height estimation of sugarcane using an unmanned aerial system (UAS) based on structure from motion (SfM) point clouds. International journal of remote sensing, 38(8-10), 2218-2230.

 

- Please improve the closing statement of your introduction, and clearly state the objective and the hypothesis of your research.

• Figure 1 can be re-organized horizontally to save space
• Line 115: What is the make/model of the 3D laser scanner? Company?
• Line 168 and line 240: That is not how you present an algorithm. The inputs and outputs are fine, however the text can be presented in blocks (as a figure), or you could use pseudo-code. 
• Tables 4, 5, 6 MUST be summarized and presented y meaningful plots. Such raw results of Mean/STD are of no interest to anyone. The manuscript can be accepted Only after the results shown in these tables have been taken care of. 
• Table 8, 9, 10: Again you have a lot of detailed results of your regression which are not necessary of interest. You could present all R2 values using dot-plots and discuss why you have achieved such high R2 values? It is not possible to have an immediate comparison of R2 values from the table (Please provide plots) and move unnecessary details to an appendix (or simply mention a summary in the text). 
• The conclusion section is poor. Half of it is more like a summary of the paper (line 461-471). Please revise and be on the point. What were the objectives, how far did you achieve the objectives? What were the limitations of the research? How can the study be improved? What are the potential applications in R&D? What are your final remarks and suggestions?
• The manuscript is suffering from a poor presentation of the results. I would suggest that the authors add more graphics to their work (both methodology and the result). 

The manuscript can be accepted after a major revision (Please note: all tables must be taken care of).

Author Response

Please see the attachment because of there are graphics.

Author Response File: Author Response.docx

Reviewer 4 Report

Major revision

This manuscript introduces High-throughput legume seed phenotyping using a handheld 3D laser scanner. In summary, the research is interesting and provides valuable results, but the current document has several weaknesses that must be strengthened in order to obtain a documentary result that is equal to the value of the publication.

General considerations:

• At the thematic level, the proposal provides a very interesting vision, automation of legume seed phenotype detection is a very useful study for modern agriculture. Nevertheless, a thorough knowledge of the detection legume seed phenotyping is not only limited to the detection time. This issue is an important limitation about the aspirations of the proposal, whose limitations should be assumed with more rigour and realism in the development of the argumentation of the manuscript.
• The document contains a total of 39 employed references, of which 22 are publications produced in the last 5 years (56.41%), 7 in the last 5-10 years (17.95%), 10 than 10 years old (25.64%) , implying a total percentage of 74.36 % recent references. In this way, the total number is sufficient, and their actuality is high.
• The automatic data collection and processing process proposed in this article, how to reflect the automation of data collection?
• It is worth affirming that the method adopted in this paper is advanced and has sufficient workload. However, compared with the most cutting-edge methods, it is less innovative and the research objects and scenarios are relatively limited.
• What is the significance of selecting 34 trait indicators in the article?Please elaborate on the application prospects and research value of 34 traits.
• The algorithm is not improved and of innovation.

Title, Abstract and Keywords:

• The abstract is complete and well structured with a good explanation of the contents of the document. Nevertheless, some parts of the abstract could better match the key words and the description of the research work below to avoid misunderstanding.

Chapter 1: Introduction

• The first paragraph introducing the research topic may present a much broad and comprehensive view of the problems related to your topicwith citations to authority references.

Cao, X.; Yan, H.; Huang, Z.; Ai, S.; Xu, Y.; Fu, R.; Zou, X. A Multi-Objective Particle Swarm Optimization for Trajectory Planning of Fruit Picking Manipulator. Agronomy 2021, 11, 2286. https://doi.org/10.3390/agronomy11112286

Wu, F.; Duan, J.; Chen, S.; Ye, Y.; Ai, P.; Yang, Z. Multi-Target Recognition of Bananas and Automatic Positioning for the Inflorescence Axis Cutting Point. Frontiers in Plant Science 2021, 12:705021. https://doi.org/ 10.3389/fpls.2021.705021

3D global mapping of large-scale unstructured orchard integrating eye-in-hand stereo vision and SLAM. Computers and Electronics in Agriculture 2021, 187: 106237. https://doi.org/10.1016/j.compag.2021.106237

• The research on phenotypic detection technology is reasonable, and the interpretation of work objectives may be effective. However, the limitations of your work are not strictly assumed and proved.
• Three dimensional reconstruction technology is applied in various engineering fields, and the scope of relevant fields should also be introduced comprehensively. When introducing the research topic, you can put forward a broader and comprehensive point of view.
• The novelty of this study is not obvious enough. In the introduction, please put your work together with your previous work in the same field and emphasize the contribution of your work.
• On a general level, the study of the proposeddetection techniques is reasonable, and the explanation of the objectives of the work may be valid. However, the limitations of your work are not rigorously assumed and justified.
• The general description of manual detection and two-dimensional image processing in the first and second paragraphs of the introduction can be more concise.
• In the third and fourth paragraphs of the introduction, whether the research of peers is fully understood and whether the description of the research status in this field is objective.
• Vision technologyapplications in various engineering fields, should also be introduced for a full glance of the scope of related area. For 3D perception, please pay attention to Three-dimensional perception of orchard banana central stock enhanced by adaptive multi-vision technology; For object detection, please refer to Color-, depth-, and shape-based 3D fruit detection.

Chapter 2: The method

• In the process of data collection, the placement of seeds may described more clearly, such as the seed placement of overlap, paste and other problems.
• In the process diagram of point cloud data adopted in Chapter 2, the scheme in Figure 2 can be more orderly and clear in the detailed explanation of the process.
• In the visualization of the morphological characteristics of a single soybean seed sample in Chapter 2, note C in Figure 7 is inconsistent with the picture content.
• Based on the plane detection method of RANSAC, what is the theoretical basis for setting the distance threshold to 0.05mm and the number of adjacent points to 15.Whether increasing the threshold can improve the segmentation accuracy.

Chapter 3: Experiments and results

• It is certain that this method has high segmentation accuracy, but it has no obvious advantage over its peers in the accuracy of 3D reconstruction in this paper.
• The description of the experimental results is too long and the experimental conclusions are not prominent.

Chapter 4: Conclusions

• After all that has been read, this technique can be considered as a seed phenotype detection technology, but it does not solve the problem of seed overlapping and pasting.
• It should mention the scope for further research as well as the implications/application of the study.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Thanks for the authors' hard work. I have no further concerns about this manuscript.

Reviewer 2 Report

The manuscript has been significantly improved

Reviewer 3 Report

The revised version has been significantly improved. The manuscript can be accepted after a final editorial check. 

Reviewer 4 Report

accept