PolSARpro is not able to read T2 data format. In the dual-pol case the C2 data format must be used !
In the dual-pol case the T11 and T22 elements of the T2 matrix will have quite the same information with T11 = |Co + Cross|^2 and T22 = |Co - Cross|^2.
Using the C2 data format, the C11 and C22 elements will be different with C11 = |Co|^2 and C22 = |Cross|^2. So it is better to use this data format.
Yes, PolSARpro has implemented the H/A/Alpha decomposition since the first version of PolSARpro in 2003 !
BUT : one must be careful when using this decomposition in the dual-pol case, as the entropy, the anisotropy and the alpha angles have NOT AT ALL the same physical information as in the fully-polarized mode.
For example, as there exist only two eigenvalues in the dual-pol case, the entropy and the anisotropy contain EXACTLY the same information !
No, it is not a good idea to complete the T2 matrix with zero values to create a pseudo T3 matrix. The reason is that the 3rd eigenvalue will be equal to 0, it follows that the anisotropy will be equal by definition to 1 and the entropy can presents NaN values due to the presence of log(0) !
I would like to calculate the shannon entropy on my dual data.
I don’t want to import raw SLC S1 data into PolSARpro because I am not familiar with PolSARpro environment yet. If there is some pre-processing user guide (I couldn’t find and inbuilt software help is under construction), kindly share if any tutorial is available.
I have applied following pre-processing on SLC product in SNAP : split>>Orb>>Cal>>deb.
I have a few doubts:
Do I need to apply: matrix>>ML>>SPK>>TC in SNAP environment and then import to PolSARpro for computation of decomposition parameters.
Entropy computation needs coherence matrix as input. And coherence matrix cant be computed by GRD product. Am I right?
Also, CAN DECOMPOSITION PARAMETERS BE APPLIED ON GRD PRODUCT?
Not necessarily. Entropy is already calculated for a given window size, so filtering before this can improve the result, but it is not mandatory. You have to check and compare. Same applies for multi-looking (will reduce the file size and processing time, if this plays a role). Terrain Correction is applied after calculation of entropy.
T matrix can only be calculated on SLC data, but I think the conversion is not necessary. If you applied polarimetric speckle filter, the output is T matrix anyway.
Hello sir,
Thankyou for your quick replies.
I would like a little clarification on the second point, if you will.
Are you saying, i dont need to do matrix generation in SNAP before importing it to POLSARpro for shannon entropy computation.
I can just do Split>orb>cal>deb>spk>ML and import to polsarpro for shannon entropy?
Does it mean that doing speckle filtering converts it into covariance matrix (T matrix)
P.S. I have always thought C2 is covariance matrix. I am confused now
sorry, I wasn’t aware that you need the Entropy from PolSARpro. There is one in SNAP as well (easier to compute), but importing data processed in SNAP is a bit tricky. I’m not sure if someone ever managed to correctly export SNAP data to the PolSARpro format and import it in PolSARpro.
On the other hand, there is a video which demonstrates these steps: https://www.youtube.com/watch?v=q79fQq9tTBg
Are you aware that you can import SLC data in PolSAR pro and still apply the SNAP preprocessing? Pottier/Hong/Yin (2019) Introduction to PolSARpro Biomass Edition
It is probably easier to process all steps directly in PolSARpro, because this grants that the data is structured correctly for your later analyses.
I again checked this, when you apply a polarimetric speckle filter on a S matrix in SNAP, the output is displayed as T-matrix .
You are right, T is coherency matrix and C is covariance matrix
source: Lee/Pottier (2009): Polarimetric radar imaging. From Basics to Applications. CRC Press.
.