SPCONVP

Description

Parameters

Parameter descriptions

Input Spectrum Library File

Specifies the name of the input spectral library file that contains spectrum records to be convolved.

Output Spectrum Library File

Specifies the name of a new spectra library file to receive the resampled spectra.

If the specified file already exists, it will be overwritten.

Spectral Library File Format

Specifies the file format of the output spectral library file.

Input Data Set

Specifies the name of the input hyperspectral data set file, which contains definitions of the target sensor for the convolved spectra.

Wavelength Interval

Specifies that the output spectra bands are to be restricted to those sensor bands whose center wavelength are either inside or outside a closed interval, specified in nanometres.

Valid Bands Only

Specifies whether the selected bands are to be restricted to those with 'plot' or 'bmask' (begin mask) quality values.

This parameter has no effect if the input file contains no band-validity metadata.

Details

SPCONVP first verifies all parameters, then extracts the sensor band response profiles from the input data set (FILE). The response profiles for all bands in the sensor definition are used, regardless of their associated band quality value or the setting of the VALONLY (Valid Bands Only) parameter. This is done because the quality of the image data stored in the data set (the data set's "band quality") is irrelevant to this operation.

SPCONVP then convolves each spectrum from the input library with the sensor band response profiles to produce a resampled spectrum where the wavelength samples align with the center wavelength values of the sensor. The resampled spectra are output to a new spectral library file (SPECFILO).

If the Valid Bands Only (VALONLY) parameter is set to "YES", only those measurement values in an input spectrum that have a "plot" or "bmask" quality value are input to the convolution operations. Otherwise, all measurement values are input.

If the output file format can store quality values, all the output spectrum samples will have an associated "plot" quality value.

The total wavelength range of the input spectra can be less that that of the union of the band response profiles. In such a case, the resampled spectra will only have measurement values for the center wavelengths of band response profiles that encompass at least one input measurement value (at least one valid input measurement value if VALONLY is "YES").

Algorithm

All the spectra in a spectrum library file have measurement value 'places' for the same set of wavelength samples. However, a given spectrum may have no measurement value in a given place. In such a case, the measurement value for that spectrum and place is null and is indicated with a negative value in that place. Different spectra may have null values in different places.

SPCONVP calculates a convolution coefficient for each wavelength sample. The convolution coefficient for a wavelength sample is simply the value of the encompassing response profile at that wavelength. A response profile encompasses a wavelength sample if the wavelength sample falls within the full-width at half maximum (FWHM) interval for that response profile.

If a wavelength sample does not fall within the FWHM interval for any response profile, then no convolution coefficient is computed for it.

Let 'n' wavelength samples be encompassed by response profile 'b'. Let the convolution coefficient for the i'th wavelength sample encompassed by response profile 'b' be denoted by C_b,i. Let the measurement value in spectrum 's' for the i'th wavelength sample encompassed by response profile 'b' be denoted by M_s,b,i.

Define NonNull(M_s,b,i) such that it returns 1 if M_s,b,i is non- null. Otherwise, it returns 0.

Denote the resampled measurement value for response profile b by R_b.

In the case where VALONLY is "NO", R_b is computed according to:

s = Sum i=1,n (C_b,i * NonNull(M_s,b,i))
a = Sum i=1,n (M_s,b,i * C_b,i * NonNull(M_s,b,i))
if (s .ne. 0) then R_b = a / s else R_b = -1

Define Valid(M_s,b,i) such that it returns 1 if M_s,b,i has a quality value of "plot" or "bmask." Otherwise, it returns 0.

In the case where VALONLY is "YES", R_b is computed according to:

s = Sum i=1,n (C_b,i * NonNull(M_s,b,i) * Valid(M_s,b,i))
a = Sum i=1,n (M_s,b,i * C_b,i * NonNull(M_s,b,i) * Valid(M_s,b,i))

if (s .ne. 0) then R_b = a / s else R_b = -1

Acknowledgements

PCI Geomatics received financial support from the Canadian Space Agency/L'Agence Spatiale Canadienne through the Earth Observation Application Development Program (EOADP) for the development of this software, under contract 9F028-0-4902/12.

References

McGraw-Hill Book Company, "Mathematical Handbook for Scientists and Engineers Second Ed."

Staenz, K., T. Szeredi, R.J. Brown, H. McNairn, R. Van Acker, "Hyperspectral Information Extraction Techniques Applied to Agricultural casi Data for Detection of Within-Field Variations".