Use VHR image to compute textural feature at the spatial scale of S2/landsat time series
So far, iota2 is not able to use jointly a very high resolution (VHR) image and a satellite image time series (SITS) at different spatial scale.
It would be interesting for users that needs high resolution feature (e.g., textural or morphological ones) to be able to:
- Extract such features from a VHR image computed at the scale of a SITS pixel,
- Include them in the processing chains (projection, concatenation with the SITS etc ...).
Feature can be selected from a pre-define list (mean, variance, min, max, Haralick feature...) or from a set of custom functions that operate on pixels.
Some additional constraints:
- The VHRS image could be a multispectral one
- The textura feature should be:
- https://www.orfeo-toolbox.org/CookBook/Applications/app_HaralickTextureExtraction.html
- https://www.orfeo-toolbox.org/CookBook-6.2/Applications/app_SFSTextureExtraction.html
- A use defined python function working on patches: pre-trained CNN, etc ...
Edited by Mathieu Fauvel