A Two-Stage Registration Strategy for Thermal–Visible Images in Substations
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsPaper title: A Two-Stage Registration Strategy for Thermal-Visible Images in Substations
This paper must be improved by providing the followings
1. Introduction: Since the paper focuses on the application to the fire monitoring, a review should include previous works published in the area.
2. Research objectives: it was difficult to understand the objectives, and subsequently it became difficult to understand how the results were evaluated against the goals.
3. Research methods: detailed rationales should be provided for each step outlined, including the details of the binary image generation, feature detection, and checkboard and Montage.
4. Experiments: Provide detailed descriptions of the images used in the experiments so that readers can understand what image features are processed.
5. Results: Provide the details and example images so that the presented results could be easily understood in comparison to the other methods.
Comments for author File: Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThis paper proposes accurate registration of images captured by tow different sensors, and the registration is conducted in two stages.
1. In the sentence of "Firstly, the OTSU threshold is calculated for the original thermal visible image, and it is converted into the corresponding binary image, which improves the calculation speed of subsequent image processing at the expense of information density. (173~175)" ==> Please explain how much information has been sacrificed in quantitatuve way? Is the loss providing no problem in practice? It would be better to quantify the loss of information.
2. In section 2.2.1, have you used other detectors rather than SIFT, SURF and Harris?
3. Before registration, the relative difference of image location can be represented using 3D rotation and translation. How did you solve this problem? Did you only consider 2D difference?
4. Only 12 manually points are used for evaluation. Is this enough?
5. Is TRE based on Euclidean distance used for accuracy evaluation. If you have used other metrics, please explain it.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe authors have presented an image registration strategy with two for thermal-visible images in substations. However there are some major concerns that needs to be addressed:
1. Firstly, the authors need to provide how they address the challenges in the existing works such as multimodal challenge which is a occurring in thermal images and parallax challenge caused due to complex geometry near to the cameras. Justify this.
2. The authors did not provide any comparison table with any existing works regarding their evaluation metrics they used. But they claimed that their model was performed well among the existing works. So, a comparison table should be added.
3. The authors need to provide the mathematical representation for the evaluation metrics they proposed for their model. With out them, it is hard to justify.
4. Regarding paper organization, the authors need to check thoroughly the entire manuscript for the errors. For instance, Figure 4 The process of constructing Gaussian pyramid by ADU-PIIFD is not provided in the manuscript but it was cited in the paper. So, the authors need to add the figure.
5. Headings are improper, for instance 2.2 is labelled initially as Proposed methodology and then labelled sub headings as 2.1 which is sort of misleading to the readers. So, the authors need to modify it.
6. The authors need to add the discussion section to address their proposed model advantages over the existing works and the limitations of their proposed model along with their future scope.
Comments on the Quality of English Language
English is to be thoroughly checked.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for AuthorsIn this paper authors present a study and proposal related to the fusion od infrared thermal imaging and visible image data for detection, identification and localization of thermal hazards, i.e. for specific application in fire detection in elecrtic grid substations. The registration of thermally visible imagese is observed and two-stage substation therma visible image registration method is proposed, in which the first step concise of binarization of original image pairs followed by the fast initial registration by an improved FAST algorithm, while in teh second step registration framework based on ADU-PIIFD method is proposed. The experimental performance estimation and coparison to other previosly proposed methods is done, which showed sustantional perfromance improvements of here proposed method. The proposed method has potential usage in different fire hazard monitoring applications.
The paper is well conceived and organized, with relatively good presentation, However, the authors are suggested to improve several aspects such as:
- The previous work in the area is presented and the state-of-the-art established. However, the more strict definition of paper contribution could be given in the introduction section. Also, the main rationale points of the proposed approach and difference to previous methods (from which many sub-steps are adopted in the proposed method) should be highlighted.
- The Figures 2 and 3 should be presented with the higher quality, while the Figure 4 is omitted in the submitted manuscript and should be inserted.
- The more precise description of the used dataset should be given.
- The choice of improtant parameters of the proposed method is just given, without the proper discussion or elaboration. Is it possible to present the result with different choice of parameters in order to present the perfromance sensitivity to this choices (or to present this in some other way).
- The choice for set of previous methods used in the performance comaprison should be discussed and elaborated.
- Have you tested SIFT, SURF and Harris method in multi-source image pair analysis for Feature Point Extraction, or the final choice is made solely on the given discussion?
- The final recheck and correction of language and formatting errors should be done.
Comments on the Quality of English LanguageThe Quality of English Language is acceptable, only the final recheck and error correction is needed.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for Authorsthank the authors for updating the manuscript as per my previous comments.
Reviewer 3 Report
Comments and Suggestions for AuthorsFrom the revised manuscript, it is observed that the authors have addressed all the concerns that are highlighted earlier.
Comments on the Quality of English Language
English is good