An Operational Processing Framework for Spaceborne SAR Formations

Round 1

Reviewer 1 Report

The authors aim in this paper to make a wavenumber domain processing scheme for Low Earth Orbiting SAR sensors hosted on small satellites, in which they attempt to generate high-resolution images by combining data acquired from each sensor. Unlike the conventional method, which is based on up-sampling and a reconstruction based on along-track FFT of the data, the paper suggests a new approach by upsampling and then focusing on each acquisition, then upscaling, allowing for distributed processing and local estimations of phases deviations and compensation.

 

The proposed method seems to be interesting. However, the authors did not compare their method to the conventional one, which raises questions about its efficiency. Therefore, I recommend that authors compare their method with the state-of-the-art and emphasize the benefits and limits of using such an approach. Below are my minor remarks:

-          Ln 29: “will come soon” -> “are expected to follow after that.”

-          Ln 36: This concept, [first] introduced: please add first

-          Ln 45: is “Each” a part of SIMO?

-          Ln 215: “expressions” -> “equations”

-          Figures 12 and 13: Please use unified labeling, either (a) and (b), or a. and b.

-          Figure 13 b: Please remake the figure without stretching and increase the font size

 

-          Figure 12a: Range “[m]”  / Figure 9: Slant Range “[km]”

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Congratulation on your elaborated research work presented in this paper.

The purposed close formation of SAR satellites is an interesting mission design. However, high-quality SAR focusing for such a formation will be demanding. Here, your presented approach is a considerable technique. The reversed order to apply focusing first and coherent superposition last sounds logical to me. It is a reasoned way-out to overcome geometrical distortions which are unavoidable in the traditional approach. It is amazing to see the scientific sound way you combine modern techniques like NuSAR and CZT in your overall system.

Small Low Earth Orbiting (LEO) satellites gain more and more importance in spaceborne remote sensing. The authors of the paper on hand are engaged in the design of a close formation of Synthetic Aperture Radar (SAR) satellites where one satellite is transmitting and several satellites in close along-track distance are receiving. High-quality SAR focusing for such a formation is a demanding issue. The aim is to coherently superpose the SAR data, received by the particular satellites, in order to obtain a single SAR image with drastically better resolution than it was available from the SAR data of the individual satellites. A considerable technique to solve these demands is topic of the paper in review.

The authors present a novel approach superior to former methods discussed for similar satellite formations in literature. The authors interchange the order of coherent superposition and SAR focusing and apply the focusing first. In this way, they are able to focus the SAR raw data from each satellite in an optimum way, considering the individual acquisition geometries of each sensor. In the process, they apply Chirp Z Transform (CZT) and the powerful but up to now just rarely used technique of Numeric SAR (NuSAR). In this way, they obtain sub-images with significantly less geometric distortions to be used as input for the coherent superposition.

The design of the SAR satellite formation in view is considerable. The presented focusing approach combines several modern techniques in a reasoned way to a novel overall system.

Your paper is well structured. The way of thoughts in the design and in the presentation of the algorithm is straight forward and scientifically sound. Elaborately designed simulations prove the quality parameters for the obtainable SAR images and the simulation results are thoroughly discussed. The drawn conclusions are supported by the research results.

Language and style in your paper are clear and concise. English usage is correct. Number and choice of cited references are adequate and indicate your very good overview on basic works as well as on most recent papers in the research area of SAR focusing.

The quality of this paper is high-class. Thus, I can recommend without any reservations the publication of your paper as is. It is a valuable contribution to the scientific discussion on the area of SAR.

Author Response

We thank this reviewer for the report and for getting straight into the indended paper aims.

Reviewer 3 Report

This paper proposes a new operational processing framework for the LEO-SAR constellations system, aiming at the HRWS imaging mode. The traditional paradigm is reversed in this work to reveal new insight and derive a new framework with a specific consideration regarding the curved orbit and phase preservation. The presentation is articulated and thorough. The major drawback is the description of the motion, especially the contents in lines 90-95. It's strongly suggested that authors give more detailed and understandable explanations about the intention of the framework. How the current processing framework is neither correct nor desired in many real scenarios? Examples are needed, and how the proposed approach can address this issue?

Regarding the experiments of the distributed scene, the scene is too simplified than those in other literature. Also, the ambiguous energy of the proposed method seems to be at -30db, much higher than those in the point-target case; what are the reasons? Besides, I doubt that such ambiguities would not cause interferences in the practical scenario.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 4 Report

The proposed paper belongs to the interesting area of SAR processing (SIMO in particular) with constellations  of small satellites. Main problems of this manuscript  are:

1. No experimental results, only some exemplary computer simulations are presented.

2. The advantages of the proposed method (Figure 2 (b) ) over the conventional method (Figure 2 (a)) are not clear.

3. Some operational  problems are ignored. They are: a) need to maintain a separation of the order of one km between satellites in the same orbit to prevent risk of collisions; b) exceeding idealization of orbits, supposed all equal and Keplerian, i.e. ignoring the various perturbations (sun-moon gravitation field, disomogeneity of the Earth's gravitation potential, residuous drag at 500 km, radiation pressure...), c) difficulty to control the positions of the phase centres of the satellites antennas at a fraction of the X-band wavelenght  (i.e.millimeters) resulting from b) and from the A&OC system errors.

4. Minor problems are  : a)Table 1 is a mix of a Table of parameters and a List of Symbols. Better to split it  in two ; b) Fig. 1 b) is unclear , shall be better explained; c) The writings within Figures 1,2,4, 6 , 7 and A15 are too small, hard to be read, d) some References are incomplete (and some have words missing) : 5., 7., 15., 36.; e) some symbols are not defined in the right place, such as tau in Fig. 1, RGC and SLC in Figure 2, f) the meaning (and aim) of some parameters is not clear as "k" in Equation (1).

 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 5 Report

Title : An operational processing framework for space-borne SAR formations

The paper investigates the potential of reversing the processing scheme for distributed SAR

systems operated with a PRF under the Nyquist criteria. It is an interesting topic for the

community considering the current trends towards distributed small satellite SAR constellations

instead of a heavy, expensive and powerful single platform systems.

My main comments on the paper are on the fact that

1) the novelty of the work is not presented in a clear way and the repetition of the

literature allocates many pages in the prepared draft

2) the advantage of the reversed processing scheme is not clarified properly.

Going into a little bit of details to the itemized comments above, some details on the state-ofthe-

art processing schemes and the one proposed in the paper (Fig.2) are defined in the

introduction, which provides enough background for the rest of the paper. However, I found Fig.

4 very confusing because it doesn’t serve for any arguments which later revisited with the

simulations. In the Section 2, the azimuth reconstruction algorithm is discussed and mentioned

that (also in the introduction) that the azimuth reconstruction is usually done in wavenumber

domain which is not correct. Most of the literature suggests to do the reconstruction in the

range-Doppler domain to compensate for the range dependency. It is clear on the other hand

that, if done in wavenumber domain range-frequency dependent corrections can be applied at

the cost of a processing in blocks. I also find the Eq. 3 confusing since it reduces

comprehensibility of the paper. It is neither the common method adopted in literature nor the

proposed by the authors. I understand that the purpose of Fig.4 and Eq. 3 is to refer to another

possible approach to reverse the processing scheme, however, this information is not presented

in a clear and understandable manner.

Section 3 explains the wavenumber-domain SAR focusing algorithms, one being more efficient

yet adopting an approximation and one being very precise but computationally expensive. Since

the system in hand is bistatic, the adaptation of these method to a bistatic system is addressed

in the Appendix. From my point of view, the information accessible in the literature is explained

very long, which should be only a short-summary and if any, the novelty of the SAR focusing

scheme should be highlighted. I can’t really say if any part of Section is not known by the

community. To my knowledge, the numerical implementation of the Omega-K algorithm and its

extension to the bistatic systems is known. Please comment on this and try to keep the section

as concise as possible and yet pointing out the novelties.

The most interesting part of the paper is the execution of the azimuth reconstruction on the

upsampled data. As it is known, by solving the linear systems to reconstruct the Doppler

bandwidth, one gets N reconstructed frequency bins for an N satellite constellation. How is it

done in the up-sampled input data? I think the section on the suggested azimuth reconstruction

has to be explained better. In the current version, the information provided on this topic is very

limited. I also don´t understand how the filters are computed if the data is focused with a

bistatic focusing kernel. Are all data channels focused with a common processing kernel, so that

there is still a phase deviation between the receive channels? I would suggest to reduce the

information on the focusing kernel by focusing on the important information, namely, the

numerical expression of the hodographs and the phase deviation/hodograph deviation between

the channels, and how the reconstruction is done after the focusing.

In the results sections, it is shown that both of the approaches agree with the results of the

state-of-the-art. However, I would appreciate more information on the advantage of choosing

the proposed technique over the one in literature and seeing some results regarding to these

points. It is roughly hinted in Page 3 between lines 90-94, but I would expect to see the results

of this claim. It seems to me that this approach actually introduces more operational load since

all of the data have to go through SAR focusing chain and then the reconstruction filters have to

be applied on the up-sampled data. These facts increase the computational complexity. It is not

so clear for me what the real advantage of this approach that’s worth to increase the

computational complexity to a much higher level.

In Figure 11, it is stated that IRF are almost the same. I see that the plot on the left has slight

shift and on the right some resolution loss. Can you comment on that? The plots of Fig. 12

should be larger. In the current version, it is not easy to evaluate the results. The resolution is

also not very good once it is zoomed in.

To summarize it again, my main impression is that the important information is not presented in

detail, the repetition of the literature takes up a lot of space and the motivation of the paper

(actual advantage of it) is not well addressed.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have addressed all my concerns. I recommend that this paper is accepted in its current form.

Author Response

Thanks for your factive help in improving the paper.

Reviewer 4 Report

The additions-corrections to the first version of this paper did not improve it enough to make it acceptable; on the contrary, the overall sloppiness of the manuscript and of the images seems to be possibly increased. This Reviewer has no enough time for point corrections everywhere , and some comment/corrections below  refer only to some sample pages. The recommendation is :  please find a careful  and available Colleague who know English well and who  is aware of overall Radar techniques and processing and possibly of SAR processing, and ask him to indicate all needed corrections (they are many) , hence suppress  this unacceptable sloppiness. After that, this manuscript may be seriously considered for publication.

Examples (first digits : number of row)

12- PRR: not defined (pulse repetition rate shall better called pulse repetition interval - PRI)

14 and 17: up-sampling and upsampling, please decide

19 - LEO not defined ( Low Earth Orbit)

31 - sub-meter means nothing, correct is: sub-meter range

32- Swaths -> swath 

33- limiting-> increasing

36 - Ref. [3] seems a wrong one

50 - DPCA is not defined (displaced phase centre antenna)

92 and 94 : SLC and RGC are not explained

213 - one reads "Error ! Ref. not found"

325 - "a small complex noise was added" : what is meant ? 

and more....

Moreover :

a) rows 45-46: it is impossible thet sensors in different orbital positions follow the same track referred to the Earth for the simple rationale that the Earth is rotating

b) in spite of the (sloppy) corrections Figure 1 remains unclear (not to mention the repeated writing Deltatau=...). Moreover Fig 1 does not explain Eqn.(1) as alelged in row 52

c) Fig. 2 is made of four parts but only two , namely (a) and (b) , are indicated, what about the more two ?

d) I cannot find in the paper a correct and needed  sentence which I've  found in the Reply letter , at its end "in order to perform the multi-channel .... spectrum". What happened to it ? 

e) about the adventages of the proposed processing method: seems it is only in the processing cost, an item which is not very significant today (and tomorrow) due to the technological progress.

Author Response

Please see the attachment

Author Response File: Author Response.pdf