Matching Confidence Constrained Bundle Adjustment for Multi-View High-Resolution Satellite Images
Round 1
Reviewer 1 Report
Dear Authors,
This manuscript proposes a method for bundle adjustment based on matching confidence of multi-view satellite images. In addition, authors provide a detailed explanation about the proposed method, information, and performance comparison with the state of the art of bundle adjustment methods.
In opinion of this reviewer, this paper is well organized and provides enough information about the topic, allowing readers to understand manuscript topics and objectives.
There are some minor revisions in my opinion should be done in the paper:
Don't repeat words of the manuscrpt's title in the keywords Please add axes names in figures 7, 8, 9, 11, 12, 13, 14 Please check references format when a web site is cited Please check for some minor english corrections or typos such as line 469Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 2 Report
Does the proposed approach is also rotation invariant? How is the generality of the proposed approach? I would suggest replacing the word “photo” with “image” in the entire manuscript. Does the defined control parameter on page 6, line 189 works for every case? Please explain how it has been determined. On page 8, line 247, how is the generality of the percentage of the highest-confidence matches? It seems that even 20% highest-confidence matches is not significantly different from 1%. Why not set it to 20% or even higher? How is the weighting function defined in equation 15? There are some standard weighting techniques exist in the literature that can be used. The robust estimation techniques might improve the results by efficiently downweighing the outliers/large residuals within the adjustment procedure. For example: Omidalizarandi, M., Kargoll, B., Paffenholz, J. A., & Neumann, I. (2019). Robust external calibration of terrestrial laser scanner and digital camera for structural monitoring. Journal of Applied Geodesy, 13(2), 105-134. Wieser, A., Brunner, F.K., 2002. Short Static GPS Sessions:Robust Estimation Results. GPS Solut. 5, 70–79.
https://doi.org/10.1007/PL00012901 The mismatch elimination procedure described on page 10, can also be improved by means of the robust estimation techniques. Therefore, might be not necessary to define a threshold of, for example, 1 pixel. The results of the proposed method in both cases of equal and geometric weights (see Figures 8 and 9) do not show a significant difference. This shows that too many matched points are discarded. It is NOT in accordance with a proper adaptive robust estimation procedure in re-weightening and outlier removal procedure. Please explain it. Figure 14 shows better results due to mismatched point elimination strategy and there should be less influence of the weightening. Please explain it.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Dear authors,
Thanks for revising and enhancing the manuscript. The responses were comprehensive and in most of the cases were quite precise. However, in one of question, it is still not clear how the t-distribution weighting were applied. But, it is not the main focus of this study. Therefore, from reviewer's point of view the paper is accepted in the present form.
Best regards,