Development of an Optical System with an Orientation Module to Detect Surface Damage to Potato Tubers
, Nikolay Sazonov and Maria GodyaevaRound 1
Reviewer 1 Report
In this work, the authors presented an efficient approach to identifying damage to potato tubers using a machine-learning method. Overall, the work is well-written and organized. However:
- The English language and presentation style should be improved. There were grammatical errors, typos, and jargon. Some parts were written verbosely and should be concise.
- Authors should highlight the main contributions of the work in the form of bullets in the last part of the introduction.
- The paper’s organization should be added in the last part of the introduction.
- Authors should clearly highlight the research gap.
- Authors should add some perspectives in the last part of the conclusion.
- References should be written in the same form and should respect the format required by the journal.
- A comparison to some related approaches is necessary.
- The English language and presentation style should be improved. There were grammatical errors, typos, and jargon. Some parts were written verbosely and should be concise.
Author Response
- The English language and presentation style should be improved. There were grammatical errors, typos, and jargon. Some parts were written verbosely and should be concise.
Reviewer № 1, Remark № 1: Notice removed.
- Authors should highlight the main contributions of the work in the form of bullets in the last part of the introduction.
Reviewer № 1, Remark № 2: Notice removed. One of the key innovations that underlies optical scanning systems is a deep learning neural network modified for agrotechnical tasks. Carrying out research on this topic will make it possible to design and create new types of devices, taking into account the characteristics of laser radiation, which are in demand in many industries and individual segments of science and technology.
- The paper’s organization should be added in the last part of the introduction.
Reviewer № 1, Remark № 3: Notice removed. This paper is organized as follows: Section 2 introduces the quality inspection and sorting system and the different components to sort food products using computer vision, described in section material and methods. Section 3 reports the experiments and results. Finally, Section 4 gives the conclusion.
- Authors should clearly highlight the research gap.
Reviewer № 1, Remark № 4: Notice removed. There is no connection between changes in parameters, collectively perceived as the "quality" of plants and hyperspectral data remains in many cases undetermined. This is exacerbated by the significant heterogeneity of the recorded parameters, which manifests itself even within the same fruit or tuber. One of the key innovations that underlies optical scanning systems is a deep learning neural network modified for agrotechnical tasks. However, to improve the quality of the optical identification system, it is necessary to ensure the development further research is needed to develop a general algorithm for defect detection of potato and upgrade the system toward real-time detection.
- Authors should add some perspectives in the last part of the conclusion.
Reviewer № 1, Remark № 5: Notice removed. The use of a vision system with a created database of models of real defects in potato tubers showed a high sorting efficiency, providing an accuracy of sorting by size by 95.4%, and an accuracy by the presence of defects by 93.1%. To improve the accuracy of sorting potato tubers using ANN, it is necessary to increase the volume of the training sample by 2-3 times.
- References should be written in the same form and should respect the format required by the journal.
Reviewer № 1, Remark № 6: Notice removed.
- A comparison to some related approaches is necessary.
Reviewer № 1, Remark № 7: To improve the accuracy of sorting potato tubers using ANN, it is necessary to increase the volume of the training sample by 2-3 times. That, in comparison with the existing research results, showing the achievement of sorting quality indicators from 89.8% to 88.2%, the developed optical identification system provides an excess in the range of values up to 7% and 6% in terms of sorting accuracy by size and presence defects, respectively.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The authors are interested in their work dealing with the problem of multi-object scanning from different angles. To ensure the proposed solution the authors have used a machine learning-based model. The authors' proposal includes the implementation of hyperspectral and RGB image processing processes from different angles. The prototype proposed in the paper
The paper is well-constructed yet some parts need to be carried out.
An overview of the state of the art is presented.
The contributions of the research are presented.
The paper is in its current need to be carried out.
· In some parts, I found long sentences which require English proofreading.
· The abstract section is too long I recommend the author check the authors' guidelines about the abstract section (usually from 150 to 200 words).
· Outlines paragraph messing in the introduction section.
· References in the paragraphs should be unified, no need for the authors to cite the names then the reference, I recommend the authors remove them from the text (example ref 3-5).
· Some parts of the paper are without any references from page 2 until the end of the paper.
· Some parts need to be carried out some of them should be flowcharts rather than text, or should produce figures with paragraphs.
· Sometimes I find two figures on one page, the authors should reorganize the paper.
· References section Some works are too old (1998, 2005,…) I recommend the authors renew some of them.
· Table 1 needs to be written in English, especially the head table.
· There are no details about the used dataset in the paper.
Author Response
- The English language and presentation style should be improved. There were grammatical errors, typos, and jargon. Some parts were written verbosely and should be concise.
Reviewer № 2, Remark № 1: Notice removed.
- The abstract section is too long I recommend the author check the authors' guidelines about the abstract section (usually from 150 to 200 words).
Reviewer № 2, Remark № 2: Notice removed.
- Outlines paragraph messing in the introduction section.
Reviewer № 2, Remark № 3: Notice removed. This paper is organized as follows: Section 2 introduces the quality inspection and sorting system and the different components to sort food products using computer vision, described in section material and methods. Section 3 reports the experiments and results. Finally, Section 4 gives the conclusion.
- References in the paragraphs should be unified, no need for the authors to cite the names then the reference, I recommend the authors remove them from the text (example ref 3-5).
Reviewer № 2, Remark № 4: Notice removed. Links are formatted in accordance with the requirements of the editors of the journal.
- Some parts of the paper are without any references from page 2 until the end of the paper.
Reviewer № 2, Remark № 5: Notice removed. Links are placed in the text of the article based on the affiliation.
- Some parts need to be carried out some of them should be flowcharts rather than text, or should produce figures with paragraphs.
Reviewer № 2, Remark № 6: Notice removed. Added a flowchart, as well as pictures and paragraphs
- Sometimes I find two figures on one page, the authors should reorganize the paper.
Reviewer № 2, Remark № 7: Notice removed.
- References section Some works are too old (1998, 2005,…) I recommend the authors renew some of them.
Reviewer № 2, Remark № 8: Notice removed.
- Table 1 needs to be written in English, especially the head table.
Reviewer № 2, Remark № 9: Notice removed.
- There are no details about the used dataset in the paper.
Reviewer № 2, Remark № 10: Notice removed. A library consisting of 3731 photographs of diseased potato tubers and 1520 photographs of infected potatoes of three varieties "Nevsky", "Red Scarlet" and "Nadezhda" was used as a training sample.
Author Response File: Author Response.pdf
Reviewer 3 Report
The authors present an interesting study of developing an optical system to identify damage in potatoes. The idea holds some novelty but the study lacks the way to show this novelty starting from the title. The study should include some numbers about the classification results obtained from the trials. Below are comments that might be useful.
ABSTRACT
No numerical values are listed to summarise the results.
Introduction
The authors must add a paragraph about the different types of damage that occur to potato tubers and the reasons associated with ach type.
The authors must state examples of previous studies that discuss the different methods of detecting damage in potatoes potato whether optical or not and the advantage of optical sensors over other types of sensors.
The authors need to clearly state the gap in research or industry that this study is investigating,
Materials and Methods
Figures 1-2: needs to be improved. It is so faded
The authors have explained their idea in details but the presentation will be better in a process-like diagram.
Results
Figures 3 and 9 along with their associated text should be moved into the materials and methods.
The idea that presented about developing an optical systems should be reflected in the title.
Discussion
No comparison between the results obtained in the study and either commercial or research systems.
Conclusions
The conclusion section is too long.
English language is ok and may be a thorough rivision would be beneficial.
Author Response
- No numerical values are listed to summarise the results.
Reviewer № 3, Remark № 1: Notice removed. The use of a vision system with a created database of models of real defects in potato tubers showed a high sorting efficiency, providing an accuracy of sorting by size by 95.4%, and an accuracy by the presence of defects by 93.1%.
- The authors must add a paragraph about the different types of damage that occur to potato tubers and the reasons associated with ach typ.
Reviewer № 3, Remark № 2: Notice removed. The increasing proportion of mechanical effects on tubers during cultivation, harvesting, and storage has determined the need to conduct breeding for increased endurance of potatoes to mechanical loads. The main factors that inflict damage to potato tubers during mechanized harvesting are the design of potato harvesters and the material from which the working bodies of the machines and operating modes are made. An important role is played by the physical and mechanical properties of tubers, which, in turn, depend on the variety, agricultural techniques of cultivation, soil structure, and climatic conditions.
- The authors must state examples of previous studies that discuss the different methods of detecting damage in potatoes potato whether optical or not and the advantage of optical sensors over other types of sensors.
Reviewer № 3, Remark № 3: Notice removed. Research results are presented on pages 2-4References in the paragraphs should be unified, no need for the authors to cite the names then the reference, I recommend the authors remove them from the text (example ref 3-5).
- The authors need to clearly state the gap in research or industry that this study is investigating.
Reviewer № 3, Remark № 4: Notice removed. There is no connection between changes in parameters collectively perceived as "quality" of plants and hyperspectral data remains in many cases uncertain. This is exacerbated by the significant heterogeneity of the recorded parameters, which manifests itself even within the same fruit or tuber.
- Figures 1-2: needs to be improved. It is so faded.
Reviewer № 3, Remark № 5: Issue persisted as color saturation was not available.
- The authors have explained their idea in details but the presentation will be better in a process-like diagram.
Reviewer № 3, Remark № 6: Notice removed. The remark has been eliminated, the work plan is presented in the flowchart of Figure 5.
- Figures 3 and 9 along with their associated text should be moved into the materials and methods.
Reviewer № 3, Remark № 7: Notice removed.
- References section Some works are too old (1998, 2005,…) I recommend the authors renew some of them.
Reviewer № 3, Remark № 8: Notice removed.
- The idea that presented about developing an optical systems should be reflected in the title.
Reviewer № 3, Remark № 9: The comment has been corrected, the title of the article has been corrected: «Optical system for identification of damage to potato tubers with orientation module»
- No comparison between the results obtained in the study and either commercial or research systems.
Reviewer № 3, Remark № 10: Notice removed.
- The conclusion section is too long.
Reviewer № 3, Remark № 11: Notice removed.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The manuscript can be accepted in the current form.
Author Response
Hello dear reviewer! Thank you for your detailed review of the manuscript!
Reviewer 3 Report
General note:
The way the authors replied to the comments is insufficient as it lacks the line no. which make it difficult to follow.
The title needs to be modified as it is not correct from the language structure.
A suggested one is: “Development of an optical system accompanied with an orientation module for identifying surface damage in potato tubers”.
Introduction: The authors must add numerical values about some examples of the systems in the industry or research that detects various types of damage in potato tubers.
Figures 2-3: The authors MUST redraw these 2 figures. The current brightness is not acceptable
Discussion: The authors still didn’t address this adequately. If you are developing a system like the one in the study you need to compare its performance to what is in the marker and in the research.
Conclusions: no modification is conducted by the authors.
Moderate English language rveision is required
Author Response
- The title needs to be modified as it is not correct from the language structure.
Reviewer № 2, Remark № 1: The note has been corrected, the title of the manuscript: « Development of an optical system accompanied with an orientation module for identifying surface damage in potato tubers».
- Introduction: The authors must add numerical values about some examples of the systems in the industry or research that detects various types of damage in potato tubers.
Reviewer № 2, Remark № 2: Notice removed. Widespread vision system for the sorting line LSP-4 (National Academy of Sciences of Belarus for agricultural mechanization) already exists and provides high efficiency of sorting apples with an accuracy of sorting fruits by size by 75.4% (Figure 1). However, one of the significant drawbacks of this system is the low accuracy of determining the presence of defects - 73.1% (Figure 1).
By using a modified Otsu thresholding algorithm (Figure 2), fresh bruises were detected at each wavelength from 710 to 810 nm in 20 nm increments, with overall detection errors of 11.7% to 14.2%. The obtained accuracy ranges from 85.00% to 95.00% for Orange, Lemon, Sweet Lime and Tomato used soft-computing techniques such as Backpropagation neural network and Probabilistic neural network (Figure 3).
- Figures 2-3: The authors MUST redraw these 2 figures. The current brightness is not acceptable.
Reviewer № 2, Remark № 3: Notice removed. Figures 2 and 3 corrected.
- Discussion: The authors still didn’t address this adequately. If you are developing a system like the one in the study you need to compare its performance to what is in the marker and in the research.
Reviewer № 2, Remark № 4: An analysis of the test results showed that the optical identification system meets the requirements of the technical specifications and ensures the high-quality performance of the technological process of identifying potato tubers as part of the processing line for post-harvest processing in terms of size and the presence of defects from mechanical damage, diseases and pests. The performance of the optical identification system as part of the line for post-harvest processing of the line is 1.2 t/h. Obviously, to improve the accuracy of recognition of defects in potato tubers, it is necessary to increase the training sample by at least 2-3 times, as well as to improve the sorting processes and further stages of post-harvest processing in comparison with the LSP-4 line shown in Figure 1.
- Conclusions: no modification is conducted by the authors.
Reviewer № 2, Remark № 5: Notice removed. The conclusion was corrected, in terms of volume.
Author Response File: Author Response.pdf
