1

Topic: dune displacement map using sentinel 2 data

hi all

i would like to quantify sand motion in a sand channel in the gobi desert on the earth using two sentinel 2 dataset.
these data are multibands and i used the infrared 10m/pixel resolution for every image.
these images are just ortorectified and i think i could quantify the displacement map.
for the frequency correlator parameters i used the frequency correlator and i used the based parameters : windows size  x initial 128 final 32
                        windows size  y initial 128 final 32

                        step x 16 step y 16

           my question is this: these parameters are correct??????????

     the displacement map is little strange but the vector field is in agreement with the slip faces of the dunes.


thanks for your help

Marco

2

Re: dune displacement map using sentinel 2 data

Hi Marco,

there is no way to tell if your parameters are good or not.
There is no such a thing as "correct" parameters. There is a range of parameters that will work, each with their pros and cons, and you'll pick the one that fits your needs. Tuning the correlation parameters is a matter of trade-off between noise, resolution, blurriness, displacement amplitude, etc...

As a guideline, the tuning of the correlation parameters depends on:
- the amplitude of the displacement your are expecting. If expected displacement is 50 pixels, don't even try a (initial) correlation window of 32 or 64. You need at least 128, and probably 256 to get the correlation to find a match.
- Imagery "texture" and noise. These affects the correlation quality, and can only be tuned with a trial-and-error approach. They entirely depend on the imagery and scene. These are the factors that will drive the tuning of the (final) correlation window size. Once you make sure that the (initial) correlation window is large enough for the displacement, try different (final) window sizes and see which one offer a good compromise between noise and blurriness.

Best,
Francois