Forums » GAMMA Processing » Offset Tracking »
fringes in displacement calculation (ALOS-PALSAR)
Added by Kaian Shahateet over 5 years ago
Hello everyone,
I am using ALOS-PALSAR images to do offset tracking to study field velocities on glaciers.
After generate the field velocities .TIF, I have lines of noises along the azimuth pathway. I attached a png of the result here.
I thought it could come from a bad window of search or step length, but even changing those combination generated the same result. Also I have checked my DEM, and seems to be ok.
Any of you guys know where those fringes come from?
Regards!
Kaian Fernandes
Replies (4)
RE: fringes in displacement calculation (ALOS-PALSAR) - Added by Thorsten Seehaus over 5 years ago
Where is your studysite and whats the projection of the DEM? It looks like artifacts caused be geocoding. The standard gamma algorithm has some problems if the sat. heading angle and the y-axis of the DEM differ by more than a few degree. However, there is a work around (attached)
| README_Gamma_demo_SCH (4.41 KB) README_Gamma_demo_SCH |
RE: fringes in displacement calculation (ALOS-PALSAR) - Added by Kaian Shahateet over 5 years ago
Thorsten, thank you for your response.
Our study-site is on James Ross Island, on Peninsular Antarctica and the projection of the island is WGS84/UTM 21. Actually, the angle between them is visible.
I will try to fix it and let you know.
Thank you very much,
Kaian Fernandes
fringes in displacement calculation (ALOS-PALSAR) - Added by Charles Werner over 5 years ago
Hello,
What is the input data you are using, SLC images?
Did you coregister the images first using areas that are not moving?
What you are seeing are not fringes, but biases in the offset estimation caused
most likely by performing offset tracking on intensity images that have been
aliased by detection (conversion to intensity) and possible multi-looking.
If possible It is very important to coregister the SLC images first using areas
that are not moving, if possible.
Then, when performing offset_tracking on the SLC images, it is important to
oversample the images if possible.
I suggest the following:
Work with SLC data if at all possible, rather than detected intensity. If you
have SLC data, oversample them in range and also azimuth by a factor of 2 in
range and azimuth before detection (conversion to intensity). If the SLC image
is not over-sampled first, you introduce aliasing into the intensity image that
leads to biases in the offset tracking that cannot be corrected without loss of
information. A partial work around for these errors is to lowpass filter the
image to reduce the aliased spectrum. This work around is for the situations
where you are given detected imagery to begin with, but if you have the SLC
images, oversampling is the way to go.
Accurate layover and shadow maps are produced by the program gc_map2, and this
program is recommended if you need these products.
Attached is a memo on offset estimation that may be helpful in understanding the
relationship between oversampling and accurate estimation of offsets.
Geocode the SLC image to a pixel spacing that is about oversampled by a factor
of 2. Then detect the image. and then measure the offsets.
Best regards,
Charles
RE: fringes in displacement calculation (ALOS-PALSAR) - Added by Kaian Shahateet over 5 years ago
Thank you Thorsten and Charles for your help.
I have refused the hypothesis that the orthorectification was generating those artifacts because the images right after the offset estimation (before geocoding) had that.
Effectively it was biases. Although, previously, I was using SLC images oversampled by a factor of 2.
Taking a look on the documentation that Charles provide me, I realized that changing the command offset_pwr_tracking by offset_pwr_tracking2 could solve the problem. And in fact it solved.
Thank you all!
Kaian Fernandes Shahateet