Defect Detection on LED Chips Based on Position Pre-Estimation and Feature Enhancement
Round 1
Reviewer 1 Report
The manuscript poses a procedure for defect detection on LED chips by using position pre-estimation and feature enhancement techniques. The authors claim that the results obtained show a very relevant improvement regarding the average time needed and the accuracy in the detection.
The document shows a number of language issues, mostly related to syntax and punctuation, which affects the comprehension of some parts of the text. It is fairly well organized, and it includes an Abstract, a first section containing an Introduction and Motivation, a second Image acquisition system section, followed by a Proposed Method, an Experimental Study and a Conclusion and Suggestion sections, finishing with the listing of References used.
As it is, in my opinion the manuscript shows some value, and it seems to describe an interesting proposal for the defect detection on LED chips.
Aiming to improve the quality of the work and without any intent to underrate neither its accuracy nor its contributions, I would like to make the following suggestions:
1. In the abstract, the absolute numerical results shown in lines 21-25 should also be related to previous the state-of-the-art ones.
2. Section 1 should be perhaps split up into 'Introduction and Motivation' and 'Previous works'/'State of the Art' sections for a better organization of the document, and the latter might be extended and improved. Additionally, the text in the section requires from a detailed English language review by a native speaker. I would also recommend to include a new 'Discussion' section before the Conclusions.
3. I think section 1 should provide graphical examples of the LED chips configuration with relevant data (sizes, etc.) and their typical defects that will be aimed for detection in the manuscript.
4. Section 2 could be considered as a sub-section of section 3 as proper image acquisition is key for the methodology. Additionally, section 3 should include more detailed information on the image capture system system: equipment specs, lighting levels, image data characteristics, etc.
5. Figure 2 is too general, and while it shows the constituting functional blocks, the relationships among them, in my opinion, are not clearly defined: some relationships are missing, and the flows are not defined at all regarding the data types being exchanged between the different blocks. Additionally, the decision blocks in the Precise Location and the Defect Identification boxes are somehow confusing to me when represented as they are.
6. I consider that the algorithm described in sub-sections 3.1.1 and 3.1.2 should be completed and made more clear to facilitate its appropriateness and effectiveness regarding the intended goals. Also, please check that all the parameters (such as PTs) are properly defined.
7. In Figure 3a, the 'original picture' does not show a full rectangle, and no information is provided about how the captured image is transformed into this one before. Also, in Figure 3d, do the points refer to the chip center or to the chip corners? Regarding the same topic, please consider rewritting the paragraph from lines 156 to 163 to make it more clear.
8. In sub-section 3.2, it is not clear how the Theta angle is determined prior to using it for the geometrical transformation.
9. In line 179, the meaning of the reference to 'otherness in different batches' is not clear to me.
10. In Figure 4a, the image seems to be a grey-level and not a color one.
11. In section 3.3.2, the modifications made to the MCC matching algorithm are not clearly explained and justified, in my opinion, regarding the basic MCC.
12. In figure 5, the lines legend seems confusing to me, as red and green lines seem to have the same description.
13. In figure 7, the vertical axis units should not be below the horizontal axis, in my opinion, but besides the left vertical axis.
14. In equations 20 and 21, the matrices are not referenced from the text where they are mentioned.
15. Please, review the label text for Figure 8.
16. In equation 22, with Kernel2' do you refer to the transposed Kernel2 matrix?
17. In line 266, please state the units of the picture size.
18. In sub-section 4.2, please explain how the dataset was built: same or different chip types, same or different manufaturer or procedure, same or different batches, ... Also, please explain the programming language used and correct some typos: perhaps 'Statics' shoudl be 'Statistics' and '32G' should read '32GB'.
19. Regarding the data shown in table 2, it would be valuable to know how these times were measured, and if they refer to processor times or actual times.
20. I think that in sub-section 4.3 more information should be provided about how the benchmarking tests were carried out, especially regarding the type and use of the datasets. The bold results shown require to be supported by the environment data to assess their validity.
21. In lines 305 and 306, I find confusing the sentence: 'template pictures and chip sizes were obtained manually', and I think some clarification must be provided on this in the description of the methodology.
22. Finally, consider using commas to format the number figures in the document, for instance '1028 x 1024' could be written as '1,028 x 1,024'.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This manuscript, the authors proposed a chip segmentation method based on position pre-estimation and a modified Normalized Correlation Coefficient matching algorithm by enhancement the feature method to accurately segment LED chips and stably detect various defects.
The comments on the manuscript are required to be addressed sincerely, which are as follows:
- In line no. 16 “normalized correlation coefficient (NCC)” Please, rewrite to “Normalized Correlation Coefficient (NCC)”.
- In line no. 49 “discrete wavelet packet decomposition (DWPD)” Please, rewrite to “Discrete Wavelet Packet Decomposition (DWPD)”
- The authors, they were successful with the results as shown as in the figure 9 and the tables 1,2 and 3 That clearly shown the proposed method is effectively detect various defects.
- Just I ask to the authors to rewrite the references. There are many references in the end of title [J] for example see ref. [1],[2], and [4]. Also, the authors forgot to write pages and year too.
- The paper needs to be rewrite the English language , there are some mistake
- Otherwise, This work presented in the paper is minor variations on a well-studied subject. The paper structure is since good and is easy to follow and understand.
- I recommend to accept the manuscript after minor revision. To enhance the quality of the manuscript.
Comments for author File: Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
In this paper the authors presented the chip segmentation method based on position pre-estimation and a modified normalized correlation coefficient algorithm, as well as feature enhancement methods. The topic is interesting and matches well for MDPI Applied Sciences journal. The paper contains meaningful review of related works. However the paper has some unclear points, and the following minor concerns.
1. The article proposes a modification of the NCC method. Therefore, it would be appropriate to describe the traditional NCC method. This will show which modifications have made it possible to achieve a gain in operating speed.
2. The modified NCC method proposed in the paper contains several steps (Fig. 2). In my opinion, the article would benefit significantly if the authors performed some ablation study. For example, it would show what gain in accuracy is given by the rotation correction or the contrasting of defects.
3. There are also typos in the paper.
- Fig. 3 — on different pages
- L. 237 - «Figure 8. P Electrode-pinhole-missing»
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
Introduction is good but chapter 2 is just two sentences and a figure.
Figure 3 has two (b) (one in bold for some reason) and no (c).
The order of sections and paragraphs seems strange to me. 4 should be discussion, Experimental study should be mentioned in 2, methodology and results should be in different sections and results must be more than one sentence.
It seems that the authors know the topic, however the structure of the paper undermines their work.
Last but important, in the compared methods, all these methods could be adapted to run with the same image size as the proposed method. This would also increase (inference? it is not defined) time and potentially accuracy. Also, accuracy should not be the only performance metric, precision, recall, specificity and sensitivity should also be taken into consideration.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 5 Report
Comments and Suggestions for Authors:
In this paper, the authors present a chip segmentation method based on position pre-estimation and a modified normalized correlation coefficient (NCC) matching algorithm, as well as feature enhancement methods. The position pre-estimation method is used to avoid the interference introduced by the disordered chip arrangement and the large missing area. By modifying the NCC algorithm, matching speed is improved by eight times compared to traditional NCC while the matching result is not affected by brightness change.
The overall writing, presentation, literature review, and experimental analysis are good. However, I have a few observations:
- In Figure 3 caption, labeling (b) comes twice, after (b) it should be (c) for column coordinate map.
- Figure 7 could be moved to section 4.
- The formats of references need to be correct. In reference 1, after the title, it’s written as [J] in almost all the references. According to the Applied Sciences format, [J] should be omitted.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The Authors have made an effort to improve the quality of the manuscript, following most of the recommendations and accepting many suggestions made by this reviewer.
Even so, I would suggest the Authors to further check the document language usage. Some of the issues that in my opinion could be addressed would be the misplacement of spaces (lines 47, 190, 232, 233, 269, 278, 279, 283, 306, and the 'Image size(pixel)' in 317), missing words/punctuation (lines 109, 177, 185, 195, 198, 201, 202, 245, 249, 301).
Section 2: The description of the image acquisition should, in my opinion, include information about the image size, resolution, depth of color and encoding.
Figure 3: The response given by the Authors to Point 5 of my first revision were very clarifying, but I think they are not evident from the figure, which without those explanations might be difficult to interpret.
Figure 11: The matching time scale / units label should be on the ordinate axis, not on the abscissa one, in my opinion.
Line 333: I think the reference to 'my laboratory' should be rephrased, either to identify the laboratory owner, or to replace the mention to 'the laboratory' or 'the Authors' laboratory.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
The manuscript is improved. It is ok for publication.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
