Hello,
I need to extract the values for GLCm texture measures (contrast, ASM, Energy etc.) can anyone tell me how can I extract the values for each texture measures using image analysis option of s1tbx?
I’m not sure I understand what you’re asking. Assuming you have a S1 GRD, first calibrate the data then from the raster menu select image analysis -> texture analysis -> GLCM. Select your window size, quantization level and the features you want.
Hello,
I created VH/VV ratio band (in dB scale) and want to include GLCM, NDVI, optical bands, elevation band etc. to my classification.
I have some questions.
1- GLCM textures are based on VH and VV bands. So, does it meaningful to classify VH/VV ratio band (in dB scale) values and VH textures or VV textures together?
2- Should I do PCA before Random Forest Classification? If yes which band’s textures should I use for PCA? VH or VV? (I think for both it takes very long time as stated below)
and It is advised HH:
should I use only VV texture?
probably, the textures of VV and VH are largely redundant, if you use a random forest classifier, the algorithm selects those bands which are most useful to discriminate your classes. If your processing capacity and time is limited, calculating textures on VV only is a legit compromise.
PCA reduces the redundancies generated by texture derivation. It only makes sense if you want to limit the amount of input data for your classifier. But as you say, it takes extremely long time to compute, so I would rather suggest to use the random forest classification and input all rasters you have. Check the “evaluate feature power set” option to get a list of variables and their contribution to the final classification result.
3 Likes
thank you very much Andreas, I will try it.
i can pm you some example python script i have used to compute this for sentinel 2 data if you want.