I want to kown the GLCM must be before Speckle Filtering,Can I do Speckle Filtering firstly,then do GLCM,Will this order affect the classification results? I will use Random Forest Classifier ,Looking forward to your reply.
If you filter your image, you are likely to remove the texture you are interested in.
Some discussions on this:
2. Apply orbit file
3.Thermal noise removal
4. Radiometric calibration
6. Speckle filtering
7. Terrain correction
8.converts bands to/from dB
This order do you think right?
depends on what you want to do with the data, especially why you need textures. I’m not sure if applying speckle filters to the textures is a good idea beacause they area already based on a larger kernel.
my idea is do GLCM firstly ,Speckle filtering is After Radiometric calibration
Radiometric calibration→Speckle filtering→Terrain correction→converts bands to/from dB
I’m doing tree classification，Texture is what I need.
that sounds good, filtering the Sigma0 band makes sense.
A tutorial on Land use classification, including tree-classification with the mask manager, based on S1 data is currently in progress. In case you are interested, please send me a private message and I’ll share the draft.
OK,I already sent it to you.
I need again of your helpful advice.
Actually, I’m working with texture analysis by GLCM with SLC IW S1 images and my idea is to analyze the difference in texture parameters considering the Intensity in linear and in dB as input for GLCM step.
My workflow is:
- TopSAR split.
- Apply orbit file.
- Thermal Noise Removal.
- Converting in dB.
- Terrain Correction.
My question is related to when it is better to do GLCM.
Do you think it is correct to put GLCM after the debursting o can I move it after the terrain correction? In the last option, could TC have an effect on the results of GLCM?
Thank you in advance.
we had a similar discussion a few days ago here: Estimating Forest Above ground biomass from sentinel imagery using gray level co_occurence matrix
I knew you had the right response to my question!
Thank you again.