spectral metadata handling

Issue #705 open
Andreas Janz created an issue

Recently we decided to store band-wise wavelength, FWHM, and bad band metadata at band-level. E.g.

Only, if that is not available, we fall back to ENVI classic metadata:

The GDAL Metadata widget, does not correctly deal with that. E.g. center wavelength is taken from band-level in nanometer units and FWHM is taken from dataset-level in micrometer units:

I would suggest to:

  1. add a column for “Bad band” status (True/False or Yes/No)
  2. Remove the Wavelength Units combo box. Changing the current unit seam to do nothing anyway.
  3. Each band can have its own wavelength (nanometers, micrometers), so I would suggest to always show nanometers.

Find attached a single-band subset of the enmap_berlin test raster for reproducing the issue.

Comments (14)

  1. Andreas Janz reporter

    @Benjamin Jakimow you may use the enmapboxprocessing.rasterreader.RasterReader to query the relevant band information:

    RasterReader.wavelength(bandNo)         # in nm
    RasterReader.fwhm(bandNo)               # in nm
    RasterReader.badBandMultiplier(bandNo)  # 1 indicates a valid band (ENVI Classic style)
    

  2. Benjamin Jakimow
    • changed status to open

    I think we first need to clearly document the way how and where wavelength information (wavelength unit, wavelengths, FWHM) are

    1. searched and 2. taken in accouunt in case of competing information
  3. Andreas Janz reporter

    I would place this in the Glossary. Here is my first attempt, that needs to be completed:
    https://enmap-box.readthedocs.io/en/latest/general/glossary.html#term-wavelength

    If I remember correctly, we wanted to search in the following locations:

    1. ENVI band domain
    2. Any other band domain
    3. ENVI dataset domain
    4. Any other dataset domain

    I will update the Glossary and expand the RasterReader methodes.

    I guess you will implement the same logic in your external QPS packages.

  4. Andreas Janz reporter

    Sidenote: I really suggest to search in the band domains first, because the dataset domains might be outdated because of prior spectral subsetting.

  5. Log in to comment