## DESCRIPTION

*i.pysptools.unmix* extracts endmembers from imagery group and performs
spectral unmixing using
pysptools.
The module is a wrapper around the **pysptools** Python library, that integrates
its functionality for
Endmember Extraction
and Spectral
Unmixing into GRASS GIS.

It requires that the Python libraries **pysptools** and **scikit-learn**
are installed.

Supported algorithms for
Endmember Extraction
are:

*NFINDR*: N-FINDR endmembers induction algorithm after Winter (1999),
that also makes use of an Automatic Target Generation Process (ATGP) (Plaza &
Chang 2006). (*Default*)
*FIPPI*: Fast Iterative Pixel Purity Index after Chang (2006)
*PPI*: Pixel Purity Index

Supported algorithms for
Spectral Unmixing
are:

*FCLS*: Fully Constrained Least Squares (FCLS): Estimates endmember
abundance per pixel with the constraint that values are non-negative and sum up
to one per pixel (*Default*)
*UCLS*: Unconstrained Least Squares (UCLS): Estimates endmember
abundance per pixel in an unconstrained way
*NNLS*: Non-negative Constrained Least Squares (NNLS): Estimates endmember
abundance per pixel with the constraint that values are non-negative

## NOTES

Number of endmembers to extract (*endmember_n*) is supposed to be lower
than the number of bands in the imagery group. Only the *PPI* method can
extract more endmembers than there are bands in the imagery group.

## EXAMPLES

# List bands
bands=`g.list type=raster pattern=lsat7_2002* separator=','`
# Create imagery group
i.group group=lsat_2002 input="$bands"
# Extract Endmembers and perform spectral unmixing using pysptools
i.pysptools.unmix input=lsat_2002 endmembers=endmembers endmember_n=5 \
output=spectrum.txt prefix=lsat_spectra --v
# Compare to result from i.spec.unmix
i.spec.unmix group=lsat7_2002 matrix=sample/spectrum.dat result=lsat7_2002_unmix \
error=lsat7_2002_unmix_err iter=lsat7_2002_unmix_iterations

## REQUIREMENTS

## REFERENCES

Chang, C.-I. 2006: A fast iterative algorithm for implementation of pixel
purity index. Geoscience and Remote Sensing Letters, IEEE, 3(1): 63-67.

Plaza, A. & Chang, C.-I. 2006: Impact of Initialization on Design of
Endmember Extraction Algorithms. Geoscience and Remote Sensing,
IEEE Transactions. 44(11): 3397-3407.

Winter, M. E. 1999: N-FINDR: an algorithm for fast autonomous spectral
end-member determination in hyperspectral data. Presented at the
Imaging Spectrometry V, Denver, CO, USA, (3753): 266-275.

## SEE ALSO

*
i.spec.unmix
*
## AUTHORS

Stefan Blumentrath,
Norwegian Institute for Nature Research (NINA), Oslo, Norway

Zofie Cimburova,
Norwegian Institute for Nature Research (NINA), Oslo, Norway