A Four Stage Image Processing Algorithm for Detecting and Counting of Bagworm, Metisa plana Walker (Lepidoptera: Psychidae)
Round 1
Reviewer 1 Report
This MS was carried to detect and count of bagworm, Metisa plana using a four stage image processing algorithm, in order to ensure proper planning of control actions to manage the occurrence and damage of M. plana. The topic is very interesting and novel for the monitoring and detection of insect populations. While there were many shortcomings and errors, which were following as:
Q1: Title: No necessary to say "robust" in the title! Suggest to change as “A four stage image processing algorithm for detection and counting of bagworm, Metisa plana Walker (Lepidoptera: Psychidae)”.
Q2: Abstract: It is not suitable to say “moderate” attack of 10% - 50%, It is not heavy damage that 50% attack? And not right to say that a bagworm attack of 10%-50% leaf damage may cause 43% yield loss. Less than 43% or more than 43%? Give a range of yield loss percents. Moreover, just an average detection accuracy of 40% and 34% at 30 cm and 50 cm camera distance respectively, this is not named as "a precise monitoring system! Furthermore, the percentage of the detection increased up to 100% at a camera distance of 30 cm in close conditions. Here, how to realize this precise monitoring from 40% to 100% detection accuracy through a deep convolutional neural network? Give the simplified introduction or explanation. The last paragraph in the Abstract, “The higher pixel count and slope correlated with the dead pupae while the lower pixel count and slope, represented the living pupae.” Why? It is not necessary to distinguish the dead or living larvae or pupae of this bagworms, it is important to monitor the population occurrence of insect pests with high accuracy.
Q3: Fig.1: Just a simple diagram, and it is not suitable for counting & determination of living dead bagworms!
Q4: Fig.2: It is hard to understand the more treated image the much more faint image! How to show the clear and unique special characteristics of this bagworms from the finally treated gausian blur filter image or the image with the color space conversion from RGB to HSV in Fig.3?
Q5: Fig.4 and Fig.5: Just morphological images? How to distinguish the morphological image of this bagworms from other insects? In field, many types of insect pests may be co-occurring, how to detect and monitor the population abundance of respective species of insect pests just using the morphological images (Fig.2 - Fig.5)?
Q6: Fig.6: This figure is no meaming with too low curve lines (two colors).
Q7: Table 1: In this study, the bagworms have seven instar stages of larvae. And it is not suitable to classify three groups with large variance in real size of different instar larvae.
Q8: Fig.7: Why just select moving objects? In field, some insect larvae may be not moving some time. And pupae all no moving! And this figure is not clear!
Q9: Table 2: The dectection rate is too low even with the height increases, and this results may be affected by the distribution pattern (horizontal or vertical) even with same population density of insects.
Q10: For the color processing performance, the effect of the damaged palm leaflet which had the same color of the bagworms led to the difficulty to detect the RoI and attributed to the wrong detection of the bagworms. What about the causes that the DL approach with Faster R-CNN can enhance the detection accuracy from 21% - 53% under 30 cm distance and 21% - 46% under 50 cm distance to 90% - 100% under 30 cm distance and 80% - 90% under 50 cm distance, respectively?
Q11: Fig.14: The significant differences were noted wrong that the columns with a (color space and 30 cm) were higher than those with b (color space and 50 cm), or the columns with c (deep learning and 30 cm) higher than those with d (deep learning and 50 cm) etc.
Q12: Table 7: The title of this table is not clear. Larvae or pupae?
Other comments and queries were directly marked in the PDF file.
Author Response
All comments have been responded accordingly. Tq
Author Response File: 
Author Response.docx
Reviewer 2 Report
Based on variance partitioning analysis you found that the plant type had a predominant effect on bacterial community structure and plants were more associated with soil free-living nitrogen-fixing bacteria than symbiotic nitrogen-fixers.
There are some questions regarding to your research:
- The main question is about generalization of your findings: as you write that plant type is the most important factor affect the nitrogen-fixing bacteria. How can you extend your results to other Karst regions with different plant types? Especially as you just worked in one region (Huanjiang County in the Guangxi autonomous region of southwestern China).
- Another question: How do you interpret the relationship between plant and bacteria? Cause and effect or causal relationship? Do bacteria types affect the plant species?
- As you write in “Introduction”: Nitrogen fixation by free-living nitrogen-fixing bacteria is significant in few types of soils. So, what is the generality of your results? Is it just important on those few types of soils?
Other comments:
M&M:
Please explain why did you PCR, T-RFLP and qPCR for nifH gene?
2.4: You used the microbial DNA extracted from soil samples and then used nifH gene. How confident are you that the primer of this gene is specific for soil free-living nitrogen-fixing bacteria?
2.7: Which kind of data was used for forward selections, variance partitioning and path analysis?
Results:
Page 8: What is “after restriction using HaeIII enzymes”?
In Fig.2: The major part of your variation is unexplained (56.1%). How do you interpret this?
Author Response
Done all correction. Tq
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
English editing of the manuscript is required. There are some examples:
Manuscript title: …detection and counting …
Abstract:
Line 3: … monitoring and detecting …
Other comments:
Abstract:
Line 10: convolutional neural network: you previously used its abbreviation (CNN) in line 9. Please mention the abbreviation in the first use then apply this in next places.
Line 11: the percentage of the detection? Is this detection accuracy?
Line 12: What is the benefit of “distinguish between the living and dead larvae? How it can be useful in a pest control and management view?
Introduction:
What is the relevance of reference (8) with your study?
Please highlight the motivation of your work in the last paragraph of Introduction section. You should show the necessity of your work.
Materials and Methods:
The materials of this study is: 1, 100 RGB images from the site at Teluk Bunot, Banting, Selangor, Malaysia. So, my question is about the generality of your results. How your results can help the control of bagworm in other places?
Please discuss about uncertainty of your model.
What was the condition of plants (oil palm) that you take the images of bagworm on them?
Author Response
Please see the attachment
Author Response File: Author Response.docx
Reviewer 2 Report
English editing of the manuscript is required. There are some examples:
Manuscript title: …detection and counting …
Abstract:
Line 3: … monitoring and detecting …
Other comments:
Abstract:
Line 10: convolutional neural network: you previously used its abbreviation (CNN) in line 9. Please mention the abbreviation in the first use then apply this in next places.
Line 11: the percentage of the detection? Is this detection accuracy?
Line 12: What is the benefit of “distinguish between the living and dead larvae? How it can be useful in a pest control and management view?
Introduction:
What is the relevance of reference (8) with your study?
Please highlight the motivation of your work in the last paragraph of Introduction section. You should show the necessity of your work.
Materials and Methods:
The materials of this study is: 1, 100 RGB images from the site at Teluk Bunot, Banting, Selangor, Malaysia. So, my question is about the generality of your results. How your results can help the control of bagworm in other places?
Please discuss about uncertainty of your model.
What was the condition of plants (oil palm) that you take the images of bagworm on them?
Author Response
Please see the attachment
Author Response File: Author Response.docx
Round 3
Reviewer 2 Report
All comments have been addressed.
Author Response
All correction done. TQ
Author Response File: Author Response.docx
