1i.maxlik(1) GRASS GIS User's Manual i.maxlik(1)
2
3
4
6 i.maxlik - Classifies the cell spectral reflectances in imagery data.
7 Classification is based on the spectral signature information generated
8 by either i.cluster, g.gui.iclass, or i.gensig.
9
11 imagery, classification, Maximum Likelihood Classification, MLC
12
14 i.maxlik
15 i.maxlik --help
16 i.maxlik group=name subgroup=name signaturefile=name output=name
17 [reject=name] [--overwrite] [--help] [--verbose] [--quiet] [--ui]
18
19 Flags:
20 --overwrite
21 Allow output files to overwrite existing files
22
23 --help
24 Print usage summary
25
26 --verbose
27 Verbose module output
28
29 --quiet
30 Quiet module output
31
32 --ui
33 Force launching GUI dialog
34
35 Parameters:
36 group=name [required]
37 Name of input imagery group
38
39 subgroup=name [required]
40 Name of input imagery subgroup
41
42 signaturefile=name [required]
43 Name of input file containing signatures
44 Generated by either i.cluster, g.gui.iclass, or i.gensig
45
46 output=name [required]
47 Name for output raster map holding classification results
48
49 reject=name
50 Name for output raster map holding reject threshold results
51
53 i.maxlik is a maximum-likelihood discriminant analysis classifier. It
54 can be used to perform the second step in either an unsupervised or a
55 supervised image classification.
56
57 Either image classification methods are performed in two steps. The
58 first step in an unsupervised image classification is performed by
59 i.cluster; the first step in a supervised classification is executed by
60 the GRASS program g.gui.iclass. In both cases, the second step in the
61 image classification procedure is performed by i.maxlik.
62
63 In an unsupervised classification, the maximum-likelihood classifier
64 uses the cluster means and covariance matrices from the i.cluster sig‐
65 nature file to determine to which category (spectral class) each cell
66 in the image has the highest probability of belonging. In a supervised
67 image classification, the maximum-likelihood classifier uses the region
68 means and covariance matrices from the spectral signature file gener‐
69 ated by g.gui.iclass, based on regions (groups of image pixels) chosen
70 by the user, to determine to which category each cell in the image has
71 the highest probability of belonging.
72
73 In either case, the raster map output by i.maxlik is a classified image
74 in which each cell has been assigned to a spectral class (i.e., a cate‐
75 gory). The spectral classes (categories) can be related to specific
76 land cover types on the ground.
77
79 The maximum-likelihood classifier assumes that the spectral signatures
80 for each class (category) in each band file are normally distributed
81 (i.e., Gaussian in nature). Algorithms, such as i.cluster,
82 g.gui.iclass, or i.gensig, however, can create signatures that are not
83 valid distributed (more likely with g.gui.iclass). If this occurs,
84 i.maxlik will reject them and display a warning message.
85
86 The signature file (signaturefile) contains the cluster and covariance
87 matrices that were calculated by the GRASS program i.cluster (or the
88 region means and covariance matrices generated by g.gui.iclass, if the
89 user runs a supervised classification). These spectral signatures are
90 what determine the categories (classes) to which image pixels will be
91 assigned during the classification process.
92
93 The optional name of a reject raster map holds the reject threshold
94 results. This is the result of a chi square test on each discriminant
95 result at various threshold levels of confidence to determine at what
96 confidence level each cell classified (categorized). It is the reject
97 threshold map layer, and contains the index to one calculated confi‐
98 dence level for each classified cell in the classified image. 16 confi‐
99 dence intervals are predefined, and the reject map is to be interpreted
100 as 1 = keep and 16 = reject. One of the possible uses for this map
101 layer is as a mask, to identify cells in the classified image that have
102 a low probability (high reject index) of being assigned to the correct
103 class.
104
106 Second part of the unsupervised classification of a LANDSAT subscene
107 (VIZ, NIR, MIR channels) in North Carolina (see i.cluster manual page
108 for the first part of the example):
109 # using here the signaturefile created by i.cluster
110 i.maxlik group=lsat7_2002 subgroup=lsat7_2002 \
111 signaturefile=sig_cluster_lsat2002 \
112 output=lsat7_2002_cluster_classes reject=lsat7_2002_cluster_reject
113 # visually check result
114 d.mon wx0
115 d.rast.leg lsat7_2002_cluster_classes
116 d.rast.leg lsat7_2002_cluster_reject
117 # see how many pixels were rejected at given levels
118 r.report lsat7_2002_cluster_reject units=k,p
119 # optionally, filter out pixels with high level of rejection
120 # here we remove pixels of at least 90% of rejection probability, i.e. categories 12-16
121 r.mapcalc "lsat7_2002_cluster_classes_filtered = \
122 if(lsat7_2002_cluster_reject <= 12, lsat7_2002_cluster_classes, null())"
123
124 RGB composite of input data
125
126 Output raster map with pixels classified (10 classes)
127
128 Output raster map with rejection probability values (pixel classifica‐
129 tion confidence levels)
130
132 Image processing and Image classification wiki pages and for historical
133 reference also the GRASS GIS 4 Image Processing manual
134
135 g.gui.iclass, i.cluster, i.gensig, i.group, i.segment, i.smap, r.kappa
136
138 Michael Shapiro, U.S.Army Construction Engineering Research Laboratory
139 Tao Wen, University of Illinois at Urbana-Champaign, Illinois
140
142 Available at: i.maxlik source code (history)
143
144 Main index | Imagery index | Topics index | Keywords index | Graphical
145 index | Full index
146
147 © 2003-2020 GRASS Development Team, GRASS GIS 7.8.5 Reference Manual
148
149
150
151GRASS 7.8.5 i.maxlik(1)