Supervised Wishart Classification - Quad-pol RADARSAT2

[laudawn]

I use Radarsat2, also get the error, so sad

[ABraun]

basically, you can still derive polarimetric decompositions and use them as features for a supervised random forest classification. It’s not based on clusters then but still on the polarimetric features.

[laudawn]

Thanks for your patient answer，I will have a try:blush:

[ABraun]

if you only have few classes, a k-means clustering might already be sufficient

[laudawn]

Another question,If I can decomposition for Sentinel-1A SLC or GRD? Sentinel-1A has only two polarizations.

[ABraun]

there are some decompositions which also work for Dual-polarized data but the results are not very good.
There are some discussions on to what extent it makes sense for Sentinel-1 data:

H-Alpha Plane Problems

@agw2 No, sorry. As I started this discussion here, I tried to use the zones in SNAP for doing a classification. Then I realized that the result of the classes didnt fit to the zones. I thought that it is a problem of the SNAP program, but it isnt. I figured out that the H/Alpha Classifier in SNAP is not intended for the SENTINEL 1 SLC Data. It is made for Quadpol Data like Radarsat. So the Problem is, that the H/Alpha Classification in SNAP based on the entropie and alpha (thats fine) but th…

H-Alpha Plane Problems

technically, you can derive entropy and alpha, yes. But the results are not very useful, especially for cross polarization. It just doesn’t give all the information needed for polarimetry: Thus, HH-VV SAR is a suitable alternative to full-polarization SAR in certain cases. Meanwhile, the extraction performance of the other two dual-polarization SARs is badly degraded due to the lack of co-polarization. Therefore, HH-HV and HV-VV SARs cannot effectively extract the scattering mechanisms in the …

Sentinel-1 SLC Polarimetric Decomposition

most of the decompositions require 4 polarizations: HH, HV, VH and VV. However, most of the Sentinel-1 data in IW mode is in only one or two polarizations: There are studies which apply decompositions on dual polarized SAR data: www.mdpi.com/2072-4292/7/6/7447/pdf You can chose the Entropy-Alpha Dual Pol Decomposition for a try. [image] It is discussed in other topics if it makes sense:

[neeturathi]

Hi I have generated T3 matrix from Radarsat-2 Quadpol data. Applying lee filter and i also want to perform the wishart classification. I am getting the same error after choosing the training data as text file generated in supervised classification. what should i do in this, Please help …Thanks in advance

[ABraun]

please post the error message. We cannot help without this information.

[neeturathi]

i also get the same error what should i do can anyone please guide in this matter

[ABraun]

I can reproduce the error

@lveci @jun_lu Could we add this to the bug list?

[neeturathi]

[neeturathi]

[jun_lu]

A JIRA ticket has been created to track the issue: https://senbox.atlassian.net/browse/SITBX-749

[jun_lu]

I’ve just tried the processing chain (refined Lee filtering => Supervised Wishart Classification Training => Supervised Wishart Classification) with RS2 full-pol data on my computer and have no problem. Here I used Refined Lee Filter in Polarimetric Speckle Filter and the output is in T3 matrix format. For Washart classification, the input should be in either the original full-pol format or in coherence matrix T3 format. But if I forget to specify the location of the “Training Data Set” file in the UI of the Supervised Wishart Classification operator, then I got the following error.

[neeturathi]

Which java version have you installed in your system.

[ABraun]

that’s strange, when I apply the polarimetric speckle filter (output is T-matrix), train three classes and apply the trained cluster centers to the trained image, I still get the error:

Could you please share your training txt so I can check if they are somehow different?
This is mine: cluster_centers.txt (1.1 KB)

[jun_lu]

Here is my training txttraining_cluster_centers.txt (2.3 KB)

[ABraun]

Did you use the Radarsat Vancouver dataset as well?

the files look similar regarding their structure but when I apply your clusters to my data, the error message persists.

[jun_lu]

openjdk version “1.8.0_232”

[jun_lu]

Yes, here is the data set I used
RS2_OK76385_PK678063_DK606752_FQ2_20080415_143807_HH_VV_HV_VH_SLC