Greetings from Greece to all OTB users,
I’m in the process of testing the phenOTB remote module, in order to understand it’s usage, and apply the tool for phenology feature extraction. At this moment, I’m interested on verifying the actual curve fitting that takes place before the computation of the metrics.
According to the documentation this can be achieved by running the module on “profile” mode.
“The application can also instead generate an output image with the same number of bands as the input image, but replacing the value of each date for each pixel by the value taken by the logistic fitting”
My system: *Ubuntu 20.1 VM
#Installed OTB (OTB-7.3.0-Linux64.run) and Built phenOTB remote module against an installed OTB
#updated OTB - python bindings
#Installed remote module against OTB
#Exported path variables to bashrc
I created a case study with a Sentinel 2 NDVI timeseries covering 10 months of data.
NDVI layers were masked out from unwanted SCL classes
OTB Gap Filling with spline interpolation was performed, to get a complete dataset.
Applied the phenOTB tool in order to evaluate the “fitted” profile versus the interpolated timeseries.
Subsequently I sampled the interpolated & fitted layers, on random points contained inside known wheat parcels, in order to compare the timeseries.
What I noticed is that the 2 NDVI curves, on any given point, are completely identical
Running the module on the params, metrics modes, gave me reasonable results, which means that the fitting function works OK in the background, but the tool seems to fail exporting the fitted profile.
Might there be any experience on the use phenOTB that could explain what am I doing wrong here?