Skip to content

i.spec.unmix

Performs Spectral mixture analysis of satellite/aerial images

i.spec.unmix group=name matrix=name [result=string] [error=name] [iter=name] [--overwrite] [--verbose] [--quiet] [--qq] [--ui]

Example:

i.spec.unmix group=name matrix=name

grass.script.run_command("i.spec.unmix", group, matrix, result=None, error=None, iter=None, overwrite=False, verbose=False, quiet=False, superquiet=False)

Example:

gs.run_command("i.spec.unmix", group="name", matrix="name")

Parameters

group=name [required]
    Name of input imagery group
matrix=name [required]
    Open Matrix file
    Matrix file containing spectral signatures
result=string
    Name of raster map prefix to hold spectral unmixing results
error=name
    Name of raster map to hold unmixing error
iter=name
    Raster map to hold number of iterations
--overwrite
    Allow output files to overwrite existing files
--help
    Print usage summary
--verbose
    Verbose module output
--quiet
    Quiet module output
--qq
    Very quiet module output
--ui
    Force launching GUI dialog

group : str, required
    Name of input imagery group
    Used as: input, group, name
matrix : str, required
    Open Matrix file
    Matrix file containing spectral signatures
    Used as: input, file, name
result : str, optional
    Name of raster map prefix to hold spectral unmixing results
error : str, optional
    Name of raster map to hold unmixing error
    Used as: output, raster, name
iter : str, optional
    Raster map to hold number of iterations
    Used as: output, raster, name
overwrite: bool, optional
    Allow output files to overwrite existing files
    Default: False
verbose: bool, optional
    Verbose module output
    Default: False
quiet: bool, optional
    Quiet module output
    Default: False
superquiet: bool, optional
    Very quiet module output
    Default: False

DESCRIPTION

i.spec.unmix is used to perform Spectral Unmixing. The result is written in percent (rounded to nearest integer).

image-altConcept of mixed pixels (Landsat example)

EXAMPLES

This example is based on the North Carolina Sample dataset.

Prior to spectral unmixing the pixel values (DN) of the Landsat scene need to be converted to reflectance values (here: using DOS1):

# rename channels or make a copy to match i.landsat.toar's input name scheme:
g.copy raster=lsat7_2002_10,lsat7_2002.1
g.copy raster=lsat7_2002_20,lsat7_2002.2
g.copy raster=lsat7_2002_30,lsat7_2002.3
g.copy raster=lsat7_2002_40,lsat7_2002.4
g.copy raster=lsat7_2002_50,lsat7_2002.5
g.copy raster=lsat7_2002_61,lsat7_2002.61
g.copy raster=lsat7_2002_62,lsat7_2002.62
g.copy raster=lsat7_2002_70,lsat7_2002.7
g.copy raster=lsat7_2002_80,lsat7_2002.8

Calculation of reflectance values from DN using DOS1 (metadata obtained from p016r035_7x20020524.met.gz):

# set computational region to first Landsat band
g.region raster=lsat7_2002_10 -p

i.landsat.toar input=lsat7_2002. output=lsat7_2002_toar. sensor=tm7 \
  method=dos1 date=2002-05-24 sun_elevation=64.7730999 \
  product_date=2004-02-12 gain=HHHLHLHHL

The resulting Landsat bands are named lsat7_2002_toar.1 .. lsat7_2002_toar.8. They are used as input for the next steps.

In order to obtain pure spectra ("endmembers") to be searched for during the spectral unmixing process later on we can either obtain them from spectral libraries (ASTER Spectral Library, USGS Spectral Library, field spectrometer, etc.) or through a PCA analysis as follows.

In order to identify pure endmembers, they are supposed to be in the corners of the PCA feature space:

i.pca -n input=lsat7_2002_toar.1,lsat7_2002_toar.2,lsat7_2002_toar.3,lsat7_2002_toar.4,lsat7_2002_toar.5,lsat7_2002_toar.7 \
         output=pca_lsat7_2002_toar
d.mon wx0
# d.correlate or use scatterplot tool in g.gui
d.correlate map=pca_lsat7_2002_toar.1,pca_lsat7_2002_toar.2

# TODO: problem: how to obtain the unprojected coordinates for the corner pixels?
# (in 1998 the xgobi software did this nicely, check today's ggobi)

Next the ASCII file (e.g. called "spectrum.dat") containing six spectra needs to be written using either spectral data from a spectral library or from the PCA analysis.

Sample content of "spectrum.dat":

# Channels: r g b i1 i2 i3
# Enter spectra linewise!
# 1. Sagebrush
# 2. Saltbush
# 3. Soil
# 4. Dry grass
#
Matrix: 4 by 6
row0:  8.87  13.14  11.71  35.85  28.26 10.54
row1: 13.59  20.12  19.61  70.66 34.82 16.35
row2: 28.26  34.82  38.27  40.1 38.27 23.7
row3: 10.54  16.35  23.7   38.98 40.1 38.98

Spectral unmixing step (requires input data to be collected in an imagery group):

i.group group=lsat7_2002_toar subgroup=lsat7_2002_toar \
  input=lsat7_2002_toar.1,lsat7_2002_toar.2,lsat7_2002_toar.3,lsat7_2002_toar.4,lsat7_2002_toar.5,lsat7_2002_toar.7

i.spec.unmix group=lsat7_2002_toar matrix=sample/spectrum.dat result=lsat7_2002_unmix \
  error=lsat7_2002_unmix_err iter=lsat7_2002_unmix_iterations

# todo: reclass to 0..100%

REFERENCES

  • Neteler, M., 1999: Spectral Mixture Analysis von Satellitendaten zur Bestimmung von Bodenbedeckungsgraden im Hinblick auf die Erosionsmodellierung. M.Sc. thesis, University of Hannover, Germany.
  • Neteler, M., D. Grasso, I. Michelazzi, L. Miori, S. Merler, and C. Furlanello, 2004. New image processing tools for GRASS. - In Proc. Free/Libre and Open Source Software for Geoinformatics: GIS-GRASS Users Conference, 12-14 Sep 2004, Bangkok, Thailand.
  • Neteler, M., D. Grasso, I. Michelazzi, L. Miori, S. Merler, and C. Furlanello, 2005. An integrated toolbox for image registration, fusion and classification. International Journal of Geoinformatics, 1(1), pp. 51-61. (PDF)

SEE ALSO

AUTHORS

Markus Neteler, University of Hannover, 1999

Mohammed Rashad (rashadkm gmail.com) (2012, update to GRASS 7)

SOURCE CODE

Available at: i.spec.unmix source code (history)
Latest change: Thursday Mar 20 21:36:57 2025 in commit 7286ecf