I have doing azimuth filtering on the two co-registered IW SLC products in SNAP 4.0 desktop. However, the result seems very weird as it produced lots of fringes in the filtered data in azimuth direction. I used the defalut parameters of the Azimuth filter.
After TOPS- coregistration,
Radar --> Interfemetric —> Azimuth filtering
I have uploaded the image.
I am wondering what could be the reason for that.
Thanks in advance.
I am having the same issue, just applied directly after Backgeocoding and ESD.
Is this a known issue?
Snap version 4 on OS X, with latest updates installed.
What are the steps you already implemented before Azimuth filtering? and what is the goal of your process?
Before azimuth filtering, I did TOPS-coregistration. The file goal of the process is to find the coherence map of two repeat pass images.
Hopefully this will be helpful
co reg…(the out put should be deramp and dmod)…> s-1 range shift
…> s-1 azimuth shift…>deburst…>subset…>infgram
formation …> Topographic removal …>Filtering Goldstein …>
those steps worked for you? I did the Backgeocoding (with dermap and demod), as well as ESD, and after that I get a similar result to the one of sohelmsc for the azimuth filtering.
Which version of snap do you use?
of course these steps are working, I’m using the last released version 4
But did you try other pairs of images?
I don’t believe azimuth filtering is needed for S1 IW.
Why do you believe so?
According to this paper it is applied.
Falahfakhri, yes I tried with different pairs, and the fact that it happend also to sohelmsc makes me believe it is either a bug or something is wrong in our workflow/parameter setting? Btw, you mentioned Azimuth and Range shift, I refer to the Spectral filters.
I am trying to process InSAR. I came to know that range shift and azimuth shift improves the co-registration and consequently coherence is improved 9correct me if I am wrong). For which, s-1 range shift followed by s-1 azimuth shift is to be performed afte co-registration as you mentioned.
I was wondering if it is user specified shift in azimuth and range direction in Enhanced Spectral Diversity. I posted screen shot also. Would you be kind enough to describe on how to determine those shifts for an image pair.
Please take a look at this article, regardless of the software they use, but take the idea of processing chain.
“Sentinel-1 IWS mode support in the GAMMA software”
If you have a lot of motion in your scene, the ESD cannot be properly estimated. For this reason, you could calculate the ESD shift on a neighbouring pair of products for the same datatakes where that scene may be stable. The ESD shifts will be displayed in the InSAR Stack Tool Window. You can then use that on the scene with the motion.
also take a look at this “Sentinel-1 Assessment of the Interferometric Wide-Swath Mode”
Thank you very much. I did see the shifts per burst being displayed in the InSAR Stack Tool window. I am only processing in one sub-swath. I suppose the values shown in the graph are the applied shifts to each burst. Am I correct?
I also have another question related to the InSAR Stack Tool Window. I do not see any thing under co-registration residuals. There is nothing mentioned in the help as well.
I understand that the quality of InSAR is dependent on the co-registration accuracy, which should be at sub-pixel level and due to TOPS mode of Sentinel-1, co-registration in azimuth direction should be of 1/1000 of pixel.
Is there any tool or method to determine the achieved co-registration accuracy?
“co-registration in azimuth direction should be of 1/1000 of pixel. Is there any tool or method to determine the achieved co-registration accuracy?” This statement was mentioned by our colleague bhwan, So is there any procedure to increase the accuracy of cor. rather than increasing the GCP. in SNAP.
I went through the NEST software and found out that it has function of exporting residual log file. Is there something like that for SNAP.
Yes, this exactly the process I wish to follow because my study area is slow motion area (Himalayan glaciers move <20 cm per day) as mentioned in the papers Falah mentioned. I am using 6 days repeat pass imagery still i get very poor coherence. These papers do mention applying spectral shift filtering after ESD. It would be very helpful if anyone can suggest me on how to improve the coherence in SNAP.
There are two solutions to improve the coherence, the first one is the temporal segmentation and the second one is spacial segmentation.
Thank you very much Falah!
I will read about the temporal and spacial segmentation you mentioned. Is there any way to perform those operations in SNAP? Or do you know any other platform for this?