Unified Framework for Ship Detection in Multi-Frequency SAR Images: A Demonstration with COSMO-SkyMed, Sentinel-1, and SAOCOM Data
Round 1
Reviewer 1 Report
1) ‘MF/MM’ and ‘MM/MF’ are used in the text. It is suggested to write in a unified way.
2) It is suggested to give a complete processing flow chart in Section 3 for easy reading.
3) According to my experience, the CFAR detector in Section 3.2 is actually an empirical method, and there is no accurate theoretical model. For example, the clutter distribution in equation (1) is difficult to give. In addition, there is no theoretical connection between the threshold in formula (2) and formula (1). Using this threshold does not guarantee that the output meets the constant false alarm rate in the statistical sense.
4) Detecting ship targets in SAR images is actually an extended target detection problem. In terms of detection theory, it is suggested that the authors refer to the literature on extended target CFAR detector.
5) What is the meaning of 10^(-x) in Table 5.
6) ‘SAO’, ‘SEN’ and ‘CSK’ in Figure 13 are not explained. These abbreviations seem to be inconsistent with the context.
Author Response
Please see attached.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
CFAR and SLA Ship Detector Cascade for Multi-Frequency SAR data
This title does not sound right. Make it more fluent and attractive
The abstract needs to be revised, as the authors need to write the full words "CFAR" and "SLA". In the abstract, quantitative values of physical properties and the accuracy of ship detections are required.
In the introduction section, lines 16 to 24 do not match the title. This paragraph must be omitted.
I suggest writing in the opening of the introduction, " SAR data have a potential for monitoring ocean surface features such as ocean surface turbulence." One of the sources of ocean surface turbulence that can be imagined in SAR data is ship wake. In this sense, there is a lot of potential for automatic ship detection in SAR images, in addition to quantifiable measurements of physical properties like length and sort and dynamic movements like speed and bearing.
Update the references as follows:
1- Marghany M. Nonlinear Ocean dynamics: Synthetic aperture radar. Elsevier; 2021 Feb 9.
2-Vachon PW, Campbell JW, Bjerkelund CA, Dobson FW, Rey MT. Ship detection by the RADARSAT SAR: Validation of detection model predictions. Canadian Journal of Remote Sensing. 1997 Mar 1;23(1):48-59.
Then the authors begin to address the mechanisms of SAR imaging the ship and other features. The authors then explain what CFAR and SLA mean. CFAR stands for "constant false alarm rate," and SLA stands for "ship length ambiguity." Then the authors address the existing algorithms, i.e., their advantages and disadvantages. The author should consider the effect of sea clutter on ship detection in SAR images.
Improve Figure 2 by including geographic location.
Line 143-145 "The algorithm serves as a pre-screening step and is followed by spatial matching to filter the product that covers the same 145 area of interest (AOI)"". It is not clear about math of algorithm!
In lines 140 to 141, what sort of filter did the authors implement? "In the first stage, SAR products are filtered on a temporal basis, considering two products as matched if their sensing period difference is below 15 minutes".
What is the solution to equation 1? There appears to be no logical link between the three equations listed in this paper.
Based on the three equations mentioned, provide a pseudo code for ship detection.
What about the impact of volume scattering on ships in Figure 5?
I recommend including the ROC figure as well and compiling with Figure 13.
I do not recommend that this paper be published at this time until the authors have completed all of the requested revisions.
Author Response
Please see attached.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
I think the authors have done excellent revisions. I highly recommended it for publication.
