Geografic-Information-System/imagery/i.gensigset/i.gensigset.html

<h2>DESCRIPTION</h2>

<em>i.gensigset</em>
is a non-interactive method for generating input into
<em><a href="i.smap.html">i.smap</a>.</em>

It is used as the first pass in the a two-pass
classification process.  It reads a raster map layer,
called the training map, which has some of the pixels or
regions already classified.  <em>i.gensigset</em> will then
extract spectral signatures from an image based on the
classification of the pixels in the training map and make
these signatures available to
<em><a href="i.smap.html">i.smap</a>.</em>

<p>
The user would then execute the GRASS program <em>
<a href="i.smap.html">i.smap</a></em> to create the
final classified map.

<h2>OPTIONS</h2>

<h3>Parameters</h3>

<dl>

<dt><b>trainingmap=</b><em>name</em> 

<dd>ground truth training map

<p>
This raster layer, supplied as input by the user, has some
of its pixels already classified, and the rest (probably
most) of the pixels unclassified.  Classified means that
the pixel has a non-zero value and unclassified means that
the pixel has a zero value.

<p>
This map must be prepared by the user in advance by using
a combination of
<em><a href="wxGUI.vdigit.html">wxGUI vector digitizer</a></em>
and 
<em><a href="v.to.rast.html">v.to.rast</a></em>,
or some other import/developement process (e.g.,
<em><a href="v.transects.html">v.transects</a>)</em>
to define the areas representative of the classes in the image.

<p>
At present, there is no fully-interactive tool specifically
designed for producing this layer.

<dt><b>group=</b><em>name</em> 

<dd>imagery group

<p>
This is the name of the group that contains the band files
which comprise the image to be analyzed. The

<em><a href="i.group.html">i.group</a></em>

command is used to construct groups of raster layers which
comprise an image.

<p>
<dt><b>subgroup=</b><em>name</em> 

<dd>subgroup containing image files


<p>
This names the subgroup within the group that selects a
subset of the bands to be analyzed. The

<em><a href="i.group.html">i.group</a></em>

command is also used to prepare this subgroup.  The
subgroup mechanism allows the user to select a subset of
all the band files that form an image.


<dt><b>signaturefile=</b><em>name</em>

<dd>resultant signature file

<p>
This is the resultant signature file (containing the means
and covariance matrices) for each class in the training map
that is associated with the band files in the subgroup
selected.

<p>

<dt><b>maxsig=</b><em>value</em> 

<dd>maximum number of sub-signatures in any class

<br>
default: 5

<p>
The spectral signatures which are produced by this program
are "mixed" signatures (see <a href="#notes">NOTES</a>).
Each signature contains one or more subsignatures
(represeting subclasses).  The algorithm in this program
starts with a maximum number of subclasses and reduces this
number to a minimal number of subclasses which are
spectrally distinct.  The user has the option to set this
starting value with this option.
</dl>


<h2>INTERACTIVE MODE</h2>

If none of the arguments are specified on the command line,
<em>i.gensigset</em> will interactively prompt for the
names of these maps and files.

<p>
It should be noted that interactive mode here only means
interactive prompting for maps and files.  It does not mean
visualization of the signatures that result from the
process.

<p>

<A NAME="notes"></a><h2>NOTES</h2>

The algorithm in <em>i.gensigset</em> determines the
parameters of a spectral class model known as a Gaussian
mixture distribution.  The parameters are estimated using
multispectral image data and a training map which labels
the class of a subset of the image pixels.  The mixture
class parameters are stored as a class signature which can
be used for subsequent segmentation (i.e., classification)
of the multispectral image.

<p>
The Gaussian mixture class is a useful model because it can
be used to describe the behavior of an information class
which contains pixels with a variety of distinct spectral
characteristics.  For example, forest, grasslands or urban
areas are examples of information classes that a user may
wish to separate in an image.  However, each of these
information classes may contain subclasses each with its
own distinctive spectral characteristic.  For example, a
forest may contain a variety of different tree species each
with its own spectral behavior.


<p>
The objective of mixture classes is to improve segmentation
performance by modeling each information class as a
probabilistic mixture with a variety of subclasses.  The
mixture class model also removes the need to perform an
initial unsupervised segmentation for the purposes of
identifying these subclasses.  However, if misclassified
samples are used in the training process, these erroneous
samples may be grouped as a separate undesired subclass.
Therefore, care should be taken to provided accurate
training data.


<p>
This clustering algorithm estimates both the number of
distinct subclasses in each class, and the spectral mean
and covariance for each subclass.  The number of subclasses
is estimated using Rissanen's minimum description length
(MDL) criteria 
[<a href="#rissanen83">1</a>].  
This criteria attempts to determine
the number of subclasses which "best" describe the data.
The approximate maximum likelihood estimates of the mean
and covariance of the subclasses are computed using the
expectation maximization (EM) algorithm 
[<a href="#dempster77">2</a>,<a href="#redner84">3</a>].  

<h2>WARNINGS</h2>

If warnings like this occur, reducing the remaining classes to 0:

<div class="code"><pre>
...
WARNING: Removed a singular subsignature number 1 (4 remain)
WARNING: Removed a singular subsignature number 1 (3 remain)
WARNING: Removed a singular subsignature number 1 (2 remain)
WARNING: Removed a singular subsignature number 1 (1 remain)
WARNING: Unreliable clustering. Try a smaller initial number of clusters
WARNING: Removed a singular subsignature number 1 (-1 remain)
WARNING: Unreliable clustering. Try a smaller initial number of clusters
Number of subclasses is 0
</pre></div>

then the user should check for:
<ul>
<li>the range of the input data should be between 0 and 100 or 255 but not
  between 0.0 and 1.0 (<em>r.info</em> and <em>r.univar</em> show the range)</li>
<li>the training areas need to contain a sufficient amount of pixels</li>
</ul>


<h2>REFERENCES</h2>

<ul>
<li><A NAME="rissanen83">J. Rissanen,</a>
"A Universal Prior for Integers and Estimation by Minimum Description Length,"
<em>Annals of Statistics,</em> vol. 11, no. 2, pp. 417-431, 1983.</li>
<li><A NAME="dempster77">A. Dempster, N. Laird and D. Rubin,</a>
"Maximum Likelihood from Incomplete Data via the EM Algorithm,"
<em>J. Roy. Statist. Soc. B,</em> vol. 39, no. 1, pp. 1-38, 1977.</li>
<li><A NAME="redner84">E. Redner and H. Walker,</a>
"Mixture Densities, Maximum Likelihood and the EM Algorithm,"
<em>SIAM Review,</em> vol. 26, no. 2, April 1984.</li>
</ul>

<h2>SEE ALSO</h2>

<em>
<a href="i.group.html">i.group</a>,
<a href="i.smap.html">i.smap</a>,
<a href="r.info.html">r.info</a>,
<a href="r.univar.html">r.univar</a>,
<a href="wxGUI.vdigit.html">wxGUI vector digitizer</a>
</em>

<h2>AUTHORS</h2>

Charles Bouman, 
School of Electrical Engineering, Purdue University
<br>
Michael Shapiro,
U.S.Army Construction Engineering Research Laboratory

<p><i>Last changed: $Date$</i>