[antig]

Hi everybody
I have a problem with point position after export.
If I calculate points with small area method using height estimation, then points are not in right place after export (positon.png). Background reflectivity map is also orthorectified. In Geocoding window is everything ok. Points are center of reflected areas like they should be. But after export they are not anymore. If I calculate points displacement with no Height estimation, points are in right place. Why that’s way?

Attachments:

You must be logged in to view attached files.

[periz]

The short answer is: you need to take a course on the subject.
The longer answer: read this https://engineering.purdue.edu/~perissin/CE597_radar_remote_sensing/10336.pdf
and also all material you find on this website, including the papers here: https://www.sarproz.com/scientific-publications/
best

[antig]

Thank You. It was very fruitful reading. If I understood correctly the positional shift caused mainly by DEM error.
In Sarproz displacement calculation totorials mainly H is also estimated. I deal with Sentinel data mainly. According to my knowledge, Sentinel is not very good dataset for dem estimation. So, does H estimation have some other purpose also (Srproz alorithms need it) or may it ignore then my external dem accuracy satisfies me?

[periz]

this is an other good reading: https://engineering.purdue.edu/~perissin/CE697_insar/MTINSAR.pdf
avoiding estimating the DEM error can cause a wrong movement estimation (because the topographic phase acts as noise on the phase values).
but you are right, Sentinel has short normal baselines, and if your DEM is good, you can assume most of the topographic phase is gone and phase residuals are small. I would not take this assumption in urban areas, but if you analyze distributed terrain, it can be a good approximation.
In Sarproz you can do what you want. You can include/exclude the height estimation as well as many other time series options/choices.
best

[antig]

thanks again periz for good paper

[antig]

sry I dont understood correctly. Does H estimation in urban areas are nessery or not? If ext. Dem is ok.

I agree that in SarProz I can do everything. But I want to do something meaningful, because I think it is not a toy.
So I have to know that philosophy is behindthose buttons that I push.

[Yuxiao]

Let me add a little bit more of my own understanding.

First of all, the shift is caused by the fact that the height of that point you try to estimate (using PSI) is not corrected. You said it is DEM error, but the better saying should be “height estimation error”.

In Sentinel-1 data, the normal baseline is very small, hence the height ambiguity is very large, thus the estimation of height will have a larger error (variance). As Dr. Perissin said, when you are working with distributed terrain and mainly focus on the movement, a comprise could be to use the external DEM and do not estimate point height.

In the urban area, since you do not have a DSM (please note the difference between DTM and DSM), then it is generally not a good idea to use the external height as the point height value. This is well explained in the reference Dr. Perissin provided.

Q

[antig]

Does SarProz have opputurnity to choose whitch plane to orthorectify exported points. Issue is that Firstly I geolocate refelctivity map. Ussualy it fits good in Google earh. After point protsessing, using H estimator for purpses that you are talked, points dont fit to reflectivity map. I have to move them by hand right place. So exlude that move to my workflow, it would be wonderful then SarProz have an opputurnity to export calculated points orthorectified to choosen plane. In my experience now are that if I don’t use H estimation then the points are orthorectified to ext dem. If H estimation is choosed then to the estimated dem.?

[Yuxiao]

If it is a common shift for ALL the points, then please check well your reference point (assuming your reflectivity map is geocoded correctly).

If the shifts for different points are different, then, as discussed above, it is highly likely a height estimation error issue. The right approach should be to estimate the point height as good as possible. In the previous discussions, those are suggestions on how to do this under varies situations.

Q

[Author]

Posts