Is there a draft documentation for the “Compute Polarimetric Parameters” tool? I’m curious about the different output products, and am trying to figure out the best way to preprocess RADARSAT-2 data to return the following:
This tool seems to require the complex data, which precludes me from performing any polarimetric speckle filtering beforehand. Would it be better to:
- Polarimetric Speckle Filter> Generate products myself using a BandMath-style function
- Generate products using “Compute Pol-Params” tool > Non-polarimetric Speckle Filter
I’m doing very similar processing on RS2 quad-pol images and faced the same dilemma. I chose to convert the scattering matrix to C3 in order to apply a polarimetric speckle filter, and then convert the C3 matrix back to a relative scattering matrix in order to compute certain polarimetric parameters like the ratios you mention. Taking advantage of the speckle filters optimized for polarimetric data seems best in my opinion.
What tool/parameters are used to obtain the relative scattering matrix from C3?
Also, for anyone who reads this in the future, I made a mistake in my original post.
What is actually required is the relative scattering matrix (HH, VV, HV, VH), not the complex products.
I’m writing the operator myself because I didn’t find one in the toolbox. There are operators to convert from one polarimetric matrix to another (for example C3toT3), but not to go back to the relative scattering matrix representation (or I missed it).
The relative scattering matrix expressed in terms of the covariance matrix elements (C_ii) can be derived as so : latex-image-1.pdf (66.6 KB). This expression seems to check out when compared with the one found in Lee and Pottier’s Polarimetric Radar Imaging (page 179), although if someone else felt like checking it I would greatly appreciate it.
The code I wrote for this operation doesn’t work yet. I tried to formulate it in the same way as the C3toT3 operator, but this is complicated for someone like me who doesn’t usually code in Java. The alternative I’ll go for Monday is simply using band maths. The equation I pointed to will have to be modified to explicitly show real and imaginary parts of S_ii. Once I have this code working I’ll upload it with the rest of my processing scripts on GitLab.
You can use the polarimetric matrix operation to convert to C3 or given another matrix T4,T3 convert to another.
By default the polarimetric operators will automatically convert a quad pol product to T3. Therefore, if you need a T3, you could directly apply the polarimetric speckle filter or decomposition without first using the matrix operator.
The polarimetric parameters operator is currently a collection of different ratios and equations from the literature. It’s intended to be used for feature extraction together with the classifiers. Different parameters require different inputs and this needs to be clean up to inform the user better.
Luis, is it possible to get the Scattering Matrix (S) with S1TBX?
HI, jvcc, I am sorry to ask you have you find the related help documentation or papers for the “compute polarimetric parameters” tool?
thanks for your help, what I need is the formula or interpolation of the “Polarimetric Parameters” tool. There are many polarimeteric radar Indexes without any interpolation, and I can not find the related help documentation or papers.
as lveci stated, this module is a collection of different kinds of parameters. Their sources are distributed among the PolSAR literature, currently without reference in the documentation, that is true. Which one do you need? We can help you search for it.
thanks for your help. I am doing a compare of different vegetation indexes to my research, so I need all vegetation indexes including radar vegetation index (RVI), radar forest degradation index (RFDI), canopy structure index (CSI), volume scattering index (VSI) and biomass index (BMI).
the second problem is the format of input dada in this tool and polaeimetric decomposition, The input to these operator can be a full polarimetric SAR product with 8 bands, i.e. I and Q bands for HH, VV, HV and VH polarizations, or covariance matrix generated by Covariance Matrix Generation operator, or coherency matrix output by Coherency Matrix Generation operator. is there any difference in the output?
thanks again for your answer.
Let’s start with the RVI: A discussion on its origin and equation can be found here: Creating Radar Vegetation Index
I don’t think so. Personally, I would always start with the original bands. Some matrix derivations are irreversible.