Mask problem

Dear Colleagues
I’v been about synergic process using S1 S2 data for LULC applications following the tutorial by Mr. Andreas Braun for combination and Fusion S1S2. At the beginning everything goes well until the process PCA which it of course depend on the collocation of two source data S1S2. My mask area doesn’t chosen automatically which makes me choose it from the ROI mask and checked on remove non-ROI pixel. If I choose anyway and press run the process continued for long time (1 or 2 hours) just writing process bar without any progress and ending. Do am I make wrong.


I created the mask after the raster product RGB composite opened in view and input parameters of the math expression correctly and did everything correctly as the tutorial do.

good job so far!
I don’t think you did something wrong. As long as you can use both images in the stack (for example in a RGB image), the input of the PCA is fine so far. But this is one of the computationally most intensive operations in SNAP and can take very long.
To speed it up you have several options:

  • don’t use all S2 bands, they are redundant anyway - so you could try with B2, B5, B8A and B12
  • digitize another smaller AOI and use it as mask to make the area smaller

Hello Mr. ABraun
Oh good but I implemented new collocated and pca before, with same area and same band without problem, it just changed speckle filter from sigma for lee. do you think that the type of the speckle and window size may be the cause of the time lag?.

sometimes it is good to close snap and start it again, only opening the last product you worked with to release some memory. The filter should not make any difference for the computation time of subsequent tasks.

I do this with new version of snap 8 it is really became fast one without lag time. I have question in the context of the synergic process of S1 ((what about other process (apply orbit- remove thermal noise-multilloking and so on) could it implemented on S1 data for getting good results concerning the quality and quantity of Sentinel-1 images.?

Great Thanks for you.

orbit state vectors and thermal noise play a minor role in this tutorial so I left them out. The orbit files increase the position accuracy of the product (which is fine enough after terrain correction) and the thermal noise strongly varies, so in case you experience strong trends or patterns you can of course remove it as well.

1 Like