Prediction of Harvest Time of Tomato Using Mask R-CNN
Round 1
Reviewer 1 Report
The use of the R-CNN mask to detect objects in photos is widely known and has been used for many years, for example, in European agriculture. Computer image analysis with the use of neural networks develops operational research, i.e. supports methods supporting decision making. It therefore makes a major contribution to the development of Agriculture 4.0. You may be tempted to carry out an economic calculation of such an undertaking. Good method, but probably expensive.
The proposed method makes it possible to determine the degree of maturity of tomatoes, but how are those suitable for harvest marked, e.g. for pickers? It is worth adding this practical aspect in the manuscript.
Author Response
Reviewer 1
The use of the R-CNN mask to detect objects in photos is widely known and has been used for many years, for example, in European agriculture. Computer image analysis with the use of neural networks develops operational research, i.e. supports methods supporting decision making. It therefore makes a major contribution to the development of Agriculture 4.0. You may be tempted to carry out an economic calculation of such an undertaking. Good method, but probably expensive.
The proposed method makes it possible to determine the degree of maturity of tomatoes, but how are those suitable for harvest marked, e.g. for pickers? It is worth adding this practical aspect in the manuscript.
【Reply】
This proposal method assumes a tomato harvesting robot. The robot uses an RGB-D camera (e.g., Depth Camera D435i) [20] to detect and locate tomatoes. An RGB-D camera is a sensor capable of acquiring RGB images and depth data.
This method detects tomatoes from input RGB images. For a harvestable tomato, the coordinates on the RGB image are obtained. The 3D position corresponding to these coordinates is then obtained using the Depth image. This tomato location information is transferred to the robot. In the case of tomatoes that are still unripe, this method will predict the tomato harvesting time and inform the farmer of this information.
【20】Depth Camera D435i , Intel® RealSense™ Technology ,
https://www.intelrealsense.com/depth-camera-d435i/ (Accessed 10 March 2022)
【Addition】5.3 Discussion Line 260-267
This proposal method assumes a tomato harvesting robot. The robot uses an RGB-D camera (e.g., Depth Camera D435i) [19] to detect and locate tomatoes. An RGB-D camera is a sensor capable of acquiring RGB images and depth data. This method detects tomatoes from input RGB images. For a harvestable tomato, the coordinates on the RGB image are obtained. The 3D position corresponding to these coordinates is then obtained using the Depth image. This tomato location information is transferred to the robot. In the case of tomatoes that are still unripe, this method will predict the tomato harvesting time and inform the farmer of this information.
Author Response File: Author Response.pdf
Reviewer 2 Report
Minagawa and Kim have presented a novel method to predict the harvest time of individual tomato fruits. The manuscript can be accepted once the authors address the following mentioned points:
- The biggest concern is that manuscript is a research article. So, it should be presented in the journal’s format including the materials and methods, results and discussion. References are not properly arranged in the manuscript.
- I suggest authors to change the title to ‘Prediction of Harvesting Time of Tomato Using Mask R-CNN’.
Line 16- What does the authors mean by each tomato? Do they mean to say ‘individual tomato’? If yes, I think it will be good to replace ‘each tomato’ with ‘individual tomato fruits’.
- Manuscript can be checked for English language (For example Line 49, 51).
- Line 53- ‘and to remove the calyx, leaves, and branches.’ Right now, the sentence gives the sense that Mask R-CNN is also used to remove the calyx, leaves, and branches. Please modify the language.
- When authors mention about novelty of the manuscript in the abstract, it will be good to highlight the gap that has been fulfilled. Do the previous studies unable to read individual tomatoes? If yes, it should be mentioned in the abstract.
- One more point, I understand that authors have worked on tomato and the employed technique is of benefit in this plant species. Can the technique is of benefit in other crops as well? I believe that it can be atleast tried in other plant species and may be of benefit. I suggest authors to discuss about this in the manuscript. If possible, it will be good to add a few lines about the same to the abstract. It may increase the readability of the paper and technique can be seen in a broader context.
I do believe that the manuscript can be accepted once the authors address the mentioned points and enrich the manuscript with the crucial information and arrange the manuscript in the proper format.
Author Response
Reviewer 2
Thank you for your
- The biggest concern is that manuscript is a research article. So, it should be presented in the journal’s format including the materials and methods, results, and discussion. References are not properly arranged in the manuscript.
【Reply】
As you indicated, we have revised the organization of the paper as follows. And We have arranged references properly. Please check it.
- Introduction
- Conventional Research
- Tomato Harvesting Time Prediction
3.1 Proposal system
3.2 Harvesting time prediction method
- Basic experience
4.1 Method of experiment
4.2 Results
- Demonstration experiment
5.1 Method of experiment
5.2 Results
5.3 Discussion
- Conclusion
- I suggest authors to change the title to ‘Prediction of Harvesting Time of Tomato Using Mask R-CNN’.
【Reply】
As you indicated, the title has been corrected.
【Before correction】
Harvesting Time Prediction for Each Tomato Detected Using Mask R-CNN
【After correction】
Prediction of Harvesting Time of Tomato Using Mask R-CNN
Line 16- What does the authors mean by each tomato? Do they mean to say ‘individual tomato’? If yes, I think it will be good to replace ‘each tomato’ with ‘individual tomato fruits’.
【Reply】
As you pointed out, we have made corrections throughout the paper.
- Manuscript can be checked for English language (For example Line 49, 51).
【Reply】
As you indicated, we have made the following corrections.
Mask R-CNN is an effective method to detect fruit bunches in agriculture fields. For example, the method can be used to determine the harvestability of fruit bunches, to judge diseases of fruit bunches, and to locate infected parts of fruit bunches.
【Before correction】
Mask R-CNN is an effective method to detect fruit quality and maturity in agriculture fields. It has been used to determine ascertain whether or not crops can be harvested by harvesting robots, to determine the crop size and maturity, and to locate the infected part of the crop.
【After correction】1. Introduction Line 51-54
Mask R-CNN is an effective method to detect fruit bunches in agriculture fields. For example, the method can be used to determine the harvestability of fruit bunches, to judge diseases of fruit bunches, and to locate infected parts of fruit bunches.
- Line 53- ‘and to remove the calyx, leaves, and branches.’ Right now, the sentence gives the sense that Mask R-CNN is also used to remove the calyx, leaves, and branches. Please modify the language.
【Reply】
In this paper, Mask R-CNN is used to detect tomato fruit bunches from input images. It also detects the calyces, leaves, and branches within the detected tomato bunches. Then, the ripeness of tomato bunches is determined using only the detected tomato bunches without the calyces, leaves, and branches.
【Before correction】
This paper presents a proposal of a method for predicting tomato harvesting time to save labor in agriculture and to support decision-making for new farmers. Mask R-CNN is used to detect tomato bunches and to remove the calyx, leaves, and branches. Then, only the detected bunch images are used to predict the tomato harvesting time.
【after correction】1. Introduction Line 54-57
This paper presents a proposal of a method for predicting tomato harvesting time to save labor in agriculture and to support decision-making for new farmers. Mask R-CNN is used to detect tomato bunches without the calyx, leaves, and branches. Then, only the detected bunch images are used to predict the tomato harvesting time.
- When authors mention about novelty of the manuscript in the abstract, it will be good to highlight the gap that has been fulfilled. Do the previous studies unable to read individual tomatoes? If yes, it should be mentioned in the abstract.
【Reply】
As you indicated, we have changed “each tomato” to “individual tomato fruits” through the whole manuscript.
- One more point, I understand that authors have worked on tomato and the employed technique is of benefit in this plant species. Can the technique is of benefit in other crops as well? I believe that it can be at least tried in other plant species and may be of benefit. I suggest authors to discuss about this in the manuscript. If possible, it will be good to add a few lines about the same to the abstract. It may increase the readability of the paper and technique can be seen in a broader context.
【Reply】
This answer is given in 5.3 Discussion. This method is applicable to fruit kinds for which ripeness can be determined using color change. Examples include apples, mangoes, pineapples, and strawberries. To apply this method to other fruits, it is necessary to determine the number of days required from raw to ripe using a fixed-point camera in advance. The reason is that the number of days required to harvest fruit varies depending on local environmental conditions.
【Addition】5.3 Discussion Line 268-273
Furthermore, this method is applicable to fruit kinds for which ripeness can be determined using color change. Examples include apples, mangoes, pineapples, and strawberries. To apply this method to other fruits, it is necessary to determine the number of days required from raw to ripe using a fixed-point camera in advance. The reason is that the number of days required to harvest fruit varies depending on local environmental conditions.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
The manuscript can be accepted in its present form.