Correlation between SAR backscatter and optical reflectance values


#1

Hi,
I am working on urban feature characterization using sentinel -1 image. I have read that there are correlation graphs for the NDVI and SAR backscatter. Is the same method feasible for the urban features(buildings,roads,railway). Can I make a correlation between the SAR backscatter values in db for the sentinel-1 data and the corresponding reflectance values of the same feature in the optical dataset ?
Kindly help,
Thanks in advance


#2

The scattering type is another in urban areas. While you compare surface scattering or volume scattering above vegetation, backscatter in urban areas is mostly caused by corner reflection and facet scattering. This often exceeds the transmitted energy (Sigma0 >> 0) so a correlation could be based on quite widely ranging values. But you can try of course.


#3

Ok Thankyou for your inputs.
Actually I have a high resolution SAR data and I was thinking of characterizing the built -up as high rise, low rise etc by calculating the dielectric constant of the rooftop and classify it accordingly. Is this a feasible approach?

The second method of characterization is of course taking into consideration the texture parameters and sigma naught values and perform a threshold algorithm for the same .
Please suggest me on this.

Kindly help,
Thanks in advance


#4

Bozh approaches would require to normalize the influence of the incidence angle and all its effects. Even with a very high resolution DSM this is hardly possible in urban areas.
How did you come up with this approach?


#5

When you say normalize the influence of the incidence angle you mean the terrain correction and subsequently calibrating it for sigma naught values right ?
I have performed the operation and the sigma naught values for my features are in good agreement with standard range of the values for the particular image.
My method is to perform a multi-resolution segmentation on the SAR image and extract the features by using the threshold algorithm.
The threshold values will be set by using statistics and texture properties.
Kindly suggest me in this,
Thanks in advance


#6

I would rather calibrate before terrain correction but if your data ranges within a reasonable range its worth a try.


#7

Yes the values of the urban features are ranging within the standard range of the RISAT -1 image.
Thankyou for your inputs.
The process of calibration is essentially converting the DN values of the image to the backscatter value right.
In the caliberation I assume that the beta naught is got as result.
The terrain correction gives the sigma naught values.
Please correect me if I am wrong.
Thanks in advance.


#8

Calibration converts Amplitude/Intensity to Sigma0, which takes into account the incidence angle (but not the one of the buildings itself).

Terrain Correction corrects for the geometric distortions in the first place, Sigma0 stays Sigma0.


#9

Oh okay.
Thankyou for the response