Three-Dimensional Coordinate Extraction Based on Radargrammetry for Single-Channel Curvilinear SAR System

Round 1

Reviewer 1 Report

Congratulations on your interesting and instructive paper.

In the past, the applicability of radargrammetry suffered from its high accuracy demands on the underlying input SAR data. This changed in the recent year with the increase in accuracy of current and upcoming SAR systems. Your paper fits well into the current scientific discussing which came up with the reassessment of radargrammetry.

Your approach to apply radargrammetry based on a set of one image in ground range geometry and one image in slant range geometry is new and original. By considering the cylindrical symmetry of the SAR data acquisition geometry in the co-registration of corresponding image features in both images, you determine the true height of the image feature above ground from its parallax. In contrast to other radargrammetry schemes, I am familiar with, this approach aims to work for the entire image content of a SAR image and is not restricted to use only few singular points (e.g. persistent scatterers).

Your paper is well structured. The introduction clearly presents the problem statement and gives an overview on state of the art approaches to solve it. The theoretical derivation of your new approach shows a straight forward train of thoughts and is scientifically sound. Results of a simulation support and demonstrate the applicability of the theoretical concept. However, I strictly recommend a prove of concept based of real SAR data. This can be done in a future add-on paper.

The number and choice of cited references is adequate and shows your good overview on the research area of radargrammetry.

In the text, moderate changed w.r.t. English usage are required. Minor mistakes in some phrases indicate uncertainties, you have w.r.t. word forms (e.g. confusion of adjective and noun or present and past participle). But in total, the text is intelligible and clear.

The approach, presented in this paper is a valuable contribution to the scientific discussion at the area of radargrammetry and worth to be published. Beyond of some minor text editing, the paper is ready to be published in present form.

In detail, I have the following comments on the text:

p. 1, l. 30: wrong form of verb: "observing scene" -> "observed scene" (The scene does not observe anything, it is observed by the sensor.)

p. 2, l. 75: remark: "not satisfactory for application": I agree with you that at the development state of 1986 radargrammetry was more of theoretical that of practical interest. I just suggest to replace the unspecific wording "not satisfactory for application" by a more specific wording like "in most use cases not satisfactory for practical application".

p. 3, l. 93/94: remark: "two Doppler equations and two range equations": In fact, one can use also more that two input images, leading to an overdetermined equation system of more that two Doppler and range equations. In this case, one needs an optimization criterion and an optimization scheme get a solution (and as additional benefit an error estimate for the coordinates). Admittedly, the number of available images depend on the abilities of the SAR system in view.

p. 3, l. 102: remark: "need to place ground control points": The need of GCPs depends on the accuracy of the flight path and the aspired accuracy of the 3-D coordinate result. You may be right, with regard to airborne SAR. However, your statement is contradictory to experiences of other scientists made with spaceborne SAR where the orbit is very accurately measurable via GNSS at the current state of art. So please, explicitly mention here that you refer to airborne SAR.

p. 4, eq. (1): This equation is valid if the acceleration vector is constant. Otherwise, there have to appear higher order terms. These terms might be hardly determinable in a practical use case. But for the sake of mathematical exactness, please explicitly mention the validity constraints of eq. (1) in the text.

p. 6, l. 183: wrong English usage: "One can be known that" This phrase do not make any sense to me: "One [i.e. a person] can be known [=renowned?]" ??? Do you want to say something like "One might be familiar with the fact that" or "It might be known that" or "It is well-known that"? Please, rephrase the sentence w.r.t. to better English usage.

p. 6, l. 195: misleading wording: "all targets will be focused at the pixel (|R_sla|, y_sla)" Hopefully, not all targets of the imaged scene will be focus at exactly the same place in the focused image. Each focused target shall appear in the image at its particular correct position, which differs from target to target. Possibly, you want to say that something like "all targets will be focused at their respective position (|R_sla|, y_sla)".

p. 6., l. 195: typo: "salve" -> "slave"

p. 9, l. 260-272: question: It is not explicitly mentioned in the text, but I (hopefully) believe that you compute the image registration adaptively on base of patches or single targets and not globally on the base of the entire image (which would be too rigidly). Please, mention in this paragraph explicitly at which basis eq. (22) is applied.

p. 9, l. 268: typo: "vary" -> "very"

p. 9., l. 275: wrong English usage (see my comment of p. 6, l. 183): "one can be known that" At p. 9., l. 275 "it is known that" might be an adequate wording.

p. 9., l. 275: wrong verb form (maybe typo): ""contain" -> "contains"

p. 9. l. 281: wrong verb form (maybe typo): "to utilized" -> "to utilize"

p. 9, l. 308: adjective instead of noun: "cylindrical symmetric" -> "cylindrical symmetry"

p. 13, l. 346: wrong English usage (maybe an artifact from text editing): "is consist" -> "consists"

p. 13, l. 347: wrong verb form: "is" -> "are"

p 13, l. 348, p. 14, l. 353 and p. 14, l. 356: punctuation: I would prefer and suggest a colon ":" instead of the full stop "." behind the names of the processing steps. A further improvement - but this is only a suggestion - might be to highlight the names of the processing steps in these three paragraphs by using both face characters (because they are used here like headlines for the respective paragraphs).

p. 16, l. 401/402: remark: "extraction errors [....] less than 5m": 5 meters do not sound very demanding to me. But I suppose, this is due to the constraints, airborne SAR is faced with. Beyond that, one has to honor that you want to register all image features and do not do any cherry picking.

p. 16, l. 405-422: recommendation: Here, you just give a summary of your recent research work as it is presented in the paper. But you shall amend it by an outlook on future investigations at this topic. In particular, a prove of concept based on real SAR data, not given in the current paper, shall be a major part of your future research work.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

1. In the introduction part, when authors try to explain the merits and drawbacks of different methods, please indicate the platform of the sensors. For example, the linear trajectory assumption is acceptable because the perturbation force is small for the quality and velocity of a satellite, but it’s not for airborne sensors. The comparison should not be carried out between different platforms, that makes no sense. For this paper, I assume the method is for airborne images because the initial height in the simulation is 8000m. Therefore, the satellite examples are not reasonable.

2. Line 62 ‘elevation’ should be ‘slant-range’. The authors seem don’t know the difference between two directions: slant range and height. SAR doesn’t image in elevation or height direction but in slant-range and azimuth directions. Please review all these in the rest of this paper. Same as Line 48, ‘in azimuth and height directions’ should be ‘in azimuth and slant-range directions’

3. Line 62 The second problem is solved many years ago. Now the resampling accuracy is very high, and both InSAR and TomoSAR are very widely used in monitoring deformation and DEM. The problem is not the ‘possible non-uniform spatial samples’, but the uncertainty from phase unwrapping.

4. In Fig.3 I see the height of the sensor is decreased to change the view angle, or the incidence angle, of the signal. Is this situation normal? This only happens when the aircraft is landing. If this situation is rare, it’s not practical, please explain.

5. In Section 3, please detail the reason that master and slave images are focused with different algorithms, and why different focusing algorithms can improve the results. What’s the difference between focusing on the slant-range plane and the ground plane. 

6. Please use some real data instead of only simulations, there are many UAV SAR data, such as from the JPL website, that can fulfill authors’ requirements.

7. The method proposed in this paper doesn’t need GCPs, but it requires higher accuracy of the position information of the sensor platform because all ground target coordinates are based on that. Please analyze the error sources and their impact on results, with the comparison of current methods such as reference [32] and [33].

8.Line138 s should be a capital letter, same as in Fig.3.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

This paper is well written and the ideas described are innovative. I don't have much feedback for this paper, except that the introduction could be beefed up with more simplistic explanation of the process.

Author Response

Thanks for your comments and recognition of our work. In the revised manuscript, the introduction has been simplified. Moreover, we have improved the readability and English writing.

 

Round 2

Reviewer 2 Report

The method proposed in this paper doesn’t need GCPs, but it requires higher accuracy of the position information of the sensor platform because all ground target coordinates are based on that. Please analyze the error sources and their impact on results, comparing current methods such as references [32] and [33] in this manuscript.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf