Prepare samples for object-base imaging analysis
Hi,
With Raffaele we are thinking about further integrations of modules for the "moringa workflow". Before thinking of implementing segmentation step directly into iota2 we thought that we could provide directly a segmentation (raster or vector) like we provide the VHR raster.
About samples management for obia, the simplest way will be to add some steps after the "merge samples by models" step :
- at first the segmentation has to be handle by title and by models to fit with the rest of the data
- the samples files
samples_region_{}_seed_{}
will be intersected with the segments as they can be used as input for further steps (zonal statistics)
Futhermore, to split the workflow between pixel analysis and object analysis, what will be the best option between : 1- create a new parameter in chain_default like "analysis mode" with default value "pixel" and add a test in the workflow :
if analysis_mode == 'pixel':
# standard chain
elif analysis_mode == 'object':
# obia chain
2- create a new section parameters (will be necessary anyway to give segmentation raster path) with 'obia_mode' boolean parameter :
if obia mode :
# obia chain
else :
# standard chain
Solution 1 seems to me clearer than solution 2...