6b8482c·Author: Arthur Eigenbrot·Closed by: Arthur Eigenbrot·2022-07-13
Description
A while back the algorithm used for matching modstates and beams in GeometricCalibration was update to use some fancy stuff from scikit-image. It turns out this this new algorithm wasn’t very good on real data.
The good news, though, is that the new algorithm has the option to pass masks for both the target and reference image. This PR leverages that fact to mask out regions of the spectra that are not near strong spectral features. Even more good news: we already had a way to identify strong spectral features in SolarGainCalibration. So we just use that and now it works on all the real data we have.
A while back the algorithm used for matching modstates and beams in
GeometricCalibration
was update to use some fancy stuff fromscikit-image
. It turns out this this new algorithm wasn’t very good on real data.The good news, though, is that the new algorithm has the option to pass masks for both the target and reference image. This PR leverages that fact to mask out regions of the spectra that are not near strong spectral features. Even more good news: we already had a way to identify strong spectral features in
SolarGainCalibration
. So we just use that and now it works on all the real data we have.