Apply smile effect correction

[zxdawn]

• enpt version: 0.19.2
• Python version: 3.10.10
• Operating System: Mac OS M2

Description

Current version of enpt sets the smile coeff to 0:

enpt/enpt/model/metadata/metadata_sensorgeo.py

Lines 175 to 178 in e45f902

	# FIXME hardcoded - DLR does not provide any smile information
	# => smile coefficients are set to zero until we have valid ones
	self.nsmile_coef = 5
	self.smile_coef = np.zeros((self.nwvl, self.nsmile_coef), dtype=float)

I see there're some smile coeffs in the METADATA.XML file:

    <smileCorrection applied="no" parametrization="polynomial">
      <VNIR>
        <bandID number="1">
          <wavelength unit="nm">418.42</wavelength>
          <coeff0 unit="nm">-2.17157E-01</coeff0>
          <coeff1 unit="nm">4.65630E-04</coeff1>
          <coeff2 unit="nm">-1.23312E-08</coeff2>
          <coeff3 unit="nm">-4.46007E-11</coeff3>
          <coeff4 unit="nm">-7.96608E-16</coeff4>
        </bandID>
        <bandID number="2">

It seems we can apply the smilecorrection by setting these data as self.nsmile_coef.

[danschef]

Currently, a smile correction is not implemented in EnPT. There were some attempts to implement it in sicor, the atmospheric correction used under the hood of EnPT but so far as I can see (here), the implementation is not complete, yet. Probably, @unbohn knows more.

[zxdawn]

Thanks @danschef. I checked the EnMAP Guide and some codes in enpt:

enpt/enpt/model/metadata/metadata_sensorgeo.py

Lines 257 to 265 in e45f902

	for bi in band_informations:
	k = np.int64(bi.attrib['Id']) - 1
	self.wvl_center[k] = float(bi.findall("CenterWavelength")[0].text)
	self.fwhm[k] = float(bi.findall("FullWidthHalfMaximum")[0].text)
	self.l_min[k] = float(bi.findall("L_min")[0].text)
	self.l_max[k] = float(bi.findall("L_max")[0].text)
	scl = bi.findall("Smile/Coefficient")
	for sc in scl:
	self.smile_coef[k, np.int64(sc.attrib['exponent'])] = float(sc.text)

enpt/enpt/model/metadata/metadata_sensorgeo.py

Lines 300 to 312 in e45f902

	def calc_smile(self):
	"""Compute smile for each EnMAP column.
	The sum in (1) is expressed as inner product of two arrays with inner dimension as the polynomial smile
	coefficients shape = (ncols, nsmile_coef) of polynomial x
	:return:
	"""
	# smile(icol, iwvl) = sum_(p=0)^(nsmile_coef-1) smile_coef[iwvl, p] * icol**p (1)
	return np.inner(
	np.array([[icol ** p for p in np.arange(self.nsmile_coef)] for icol in np.arange(self.ncols)]),
	self.smile_coef # shape = (nwvl, nsmile_coef)
	) # shape = (ncols, nwvl)

It seems smile correction is already implemented but for is_dummy_dataformat. Is it possible to apply it based on the default L1B META file?