1i.gensigset(1) Grass User's Manual i.gensigset(1)
2
6 i.gensigset - Generate statistics for i.smap from raster map layer.
7
9 imagery
10
12 i.gensigset
13 i.gensigset help
14 i.gensigset trainingmap=string group=string subgroup=string signature‐
15 file=string [maxsig=integer]
16 Parameters:
18 trainingmap=string
19 ground truth training map
20 group=string
22 imagery group
23 subgroup=string
25 subgroup containing image files
26 signaturefile=string
28 resultant signature file
29 maxsig=integer
31 maximum number of sub-signatures in any class Default: 10
32
34 i.gensigset is a non-interactive method for generating input into
35 i.smap. It is used as the first pass in the a two-pass classification
36 process. It reads a raster map layer, called the training map, which
37 has some of the pixels or regions already classified. i.gensigset will
38 then extract spectral signatures from an image based on the classifica‐
39 tion of the pixels in the training map and make these signatures avail‐
40 able to i.smap.
41 The user would then execute the GRASS program i.smap to create the
43 final classified map.
44
46 Parameters
47 trainingmap=name
48 ground truth training map
49 This raster layer, supplied as input by the user, has some of its pix‐
51 els already classified, and the rest (probably most) of the pixels
52 unclassified. Classified means that the pixel has a non-zero value and
53 unclassified means that the pixel has a zero value.
54 This map must be prepared by the user in advance. The user must use
56 r.digit, a combination of v.digit and v.to.rast, or some other
57 import/developement process (e.g., v.in.transects) to define the areas
58 representative of the classes in the image.
59 At present, there is no fully-interactive tool specifically designed
61 for producing this layer.
62 group=name
64 imagery group
65 This is the name of the group that contains the band files which com‐
67 prise the image to be analyzed. The i.group command is used to con‐
68 struct groups of raster layers which comprise an image.
69 subgroup=name
71 subgroup containing image files
72 This names the subgroup within the group that selects a subset of the
74 bands to be analyzed. The i.group command is also used to prepare this
75 subgroup. The subgroup mechanism allows the user to select a subset of
76 all the band files that form an image.
77 signaturefile=name
79 resultant signature file
80 This is the resultant signature file (containing the means and covari‐
82 ance matrices) for each class in the training map that is associated
83 with the band files in the subgroup selected.
84 maxsig=value
86 maximum number of sub-signatures in any class
87 default: 10
88 The spectral signatures which are produced by this program are "mixed"
90 signatures (see NOTES). Each signature contains one or more subsigna‐
91 tures (represeting subclasses). The algorithm in this program starts
92 with a maximum number of subclasses and reduces this number to a mini‐
93 mal number of subclasses which are spectrally distinct. The user has
94 the option to set this starting value with this option.
95
97 If none of the arguments are specified on the command line, i.gensigset
98 will interactively prompt for the names of these maps and files.
99 It should be noted that interactive mode here only means interactive
101 prompting for maps and files. It does not mean visualization of the
102 signatures that result from the process.
103
105 The algorithm in i.gensigset determines the parameters of a spectral
106 class model known as a Gaussian mixture distribution. The parameters
107 are estimated using multispectral image data and a training map which
108 labels the class of a subset of the image pixels. The mixture class
109 parameters are stored as a class signature which can be used for subse‐
110 quent segmentation (i.e., classification) of the multispectral image.
111 The Gaussian mixture class is a useful model because it can be used to
113 describe the behavior of an information class which contains pixels
114 with a variety of distinct spectral characteristics. For example, for‐
115 est, grasslands or urban areas are examples of information classes that
116 a user may wish to separate in an image. However, each of these infor‐
117 mation classes may contain subclasses each with its own distinctive
118 spectral characteristic. For example, a forest may contain a variety
119 of different tree species each with its own spectral behavior.
120 The objective of mixture classes is to improve segmentation performance
122 by modeling each information class as a probabilistic mixture with a
123 variety of subclasses. The mixture class model also removes the need
124 to perform an initial unsupervised segmentation for the purposes of
125 identifying these subclasses. However, if misclassified samples are
126 used in the training process, these erroneous samples may be grouped as
127 a separate undesired subclass. Therefore, care should be taken to pro‐
128 vided accurate training data.
129 This clustering algorithm estimates both the number of distinct sub‐
131 classes in each class, and the spectral mean and covariance for each
132 subclass. The number of subclasses is estimated using Rissanen's mini‐
133 mum description length (MDL) criteria [1]. This criteria attempts to
134 determine the number of subclasses which "best" describe the data. The
135 approximate maximum likelihood estimates of the mean and covariance of
136 the subclasses are computed using the expectation maximization (EM)
137 algorithm [3].
138
140 J. Rissanen, "A Universal Prior for Integers and Estimation by Minimum
141 Description Length," Annals of Statistics, vol. 11, no. 2, pp. 417-431,
142 1983. A. Dempster, N. Laird and D. Rubin, "Maximum Likelihood from
143 Incomplete Data via the EM Algorithm," J. Roy. Statist. Soc. B, vol.
144 39, no. 1, pp. 1-38, 1977. E. Redner and H. Walker, "Mixture Densi‐
145 ties, Maximum Likelihood and the EM Algorithm," SIAM Review, vol. 26,
146 no. 2, April 1984.
147
149 i.group for creating groups and subgroups
150 v.digit and r.digit for interactively creating the training map.
152 i.smap for creating a final classification layer from the signatures
154 generated by i.gensigset.
155
157 Charles Bouman, School of Electrical Engineering, Purdue University
158 Michael Shapiro, U.S.Army Construction Engineering Research Laboratory
159 Last changed: $Date: 2003/04/17 14:49:32 $
161 Full index
163GRASS 6.2.2 i.gensigset(1)