How Create Signatures works
The signature file generated by the Create Signatures tool is a statistical description of the classes derived from the samples identified on the input raster or feature sample data. The file consists of two sections:
- The general information for all classes, such as the number of layers, input raster names, and number of classes.
- Signature statistics for each class, which consist of the number of samples and the means and covariance matrices.
The tool creates a signature file to be used as the input for other multivariate analysis tools. For example, the Maximum Likelihood Classification tool performs a maximum likelihood classification that requires both the class mean vectors and covariance matrices from a signature file.
Example
The signature file produced by Create Signatures begins with a header, which is commented out. The header retains the inputs used to create the signature file. The class names are optional and can be entered through the Sample field or they can be added to the resulting signature file using any text editor.
Compute covariance option
The following example shows a signature file. The input is a multiband raster named redlands. The sample data is a raster redzone5, and has five sample classes. The Compute covariance matrices option has been left with the default enabled setting (COVARIANCE).
- Settings used in the Create Signatures tool dialog box:
Input raster bands : redlands
Input raster or feature sample data : redzone5
Sample field : "Value"
Output signature file : z5red.gsg
Compute covariance matrices : on
The output signature file is listed below:
# Signatures Produced by ClassSig from Zone-Grid redsamp5 and Stack redlands
# Number of selected grids
/* 3
# Layer-Number Grid-name
/* 1 redlands3
/* 2 redlands1
/* 3 redlands2
# Type Number of Classes Number of Layers Number of Parametric Layers
1 5 3 3
# ----------------------------------------------------------------------
# Class ID Number of Cells Class Name
1 654 sand
# Layers 1 2 3
# Means
170.4908 155.7569 161.9419
# Covariance
1 292.6546 182.3661 186.2583
2 182.3661 127.8076 139.3009
3 186.2583 139.3009 196.3029
# ---------------------------------------------------------------
# Class ID Number of Cells Class Name
2 585 urban
# Layers 1 2 3
# Means
104.5009 92.4410 92.0513
# Covariance
1 384.6580 552.1828 389.0496
2 552.1828 1378.6750 863.5595
3 389.0496 863.5595 772.2063
# ---------------------------------------------------------------
# Class ID Number of Cells Class Name
3 783 forest
# Layers 1 2 3
# Means
27.0026 174.3768 72.7931
# Covariance
1 241.0818 -14.6301 293.7806
2 -14.6301 764.2914 221.4054
3 293.7806 221.4054 527.0799
# ---------------------------------------------------------------
# Class ID Number of Cells Class Name
4 951 water
# Layers 1 2 3
# Means
1.1504 0.0515 0.0873
# Covariance
1 7.2753 3.9638 6.4848
2 3.9638 2.5247 4.0702
3 6.4848 4.0702 6.5724
# -----------------------------------------------------------------
# Class ID Number of Cells Class Name
5 969 agri_field
# Layers 1 2 3
# Means
32.4675 232.7781 85.4149
# Covariance
1 423.1004 -684.8693 324.1354
2 -684.8693 1271.6315 -509.0008
3 324.1354 -509.0008 366.1232Means only option
If the covariance matrices are not required, disable the Compute covariance matrices option (MEAN_ONLY). Following is the signature file from the same data as above, but without calculating the covariance matrices:
Input raster bands : redlands Input raster or feature sample data : redzone5 Sample field : "Value" Output signature file : z5red.gsg Compute covariance matrices : off
The output signature file is listed below:
# Number of selected grids
/* 3
# Layer-Number Grid-name
/* 1 redlands3
/* 2 redlands1
/* 3 redlands2
# Type Number of Classes Number of Layers Number of Parametric Layers
1 5 3 3
# ----------------------------------------------------------------------
# Class ID Number of Cells Class Name
1 654 sand
# Layers 1 2 3
# Means
170.4908 155.7569 161.9419
# ---------------------------------------------------------------
# Class ID Number of Cells Class Name
2 585 urban
# Layers 1 2 3
# Means
104.5009 92.4410 92.0513
# ---------------------------------------------------------------
# Class ID Number of Cells Class Name
3 783 forest
# Layers 1 2 3
# Means
27.0026 174.3768 72.7931
# ---------------------------------------------------------------
# Class ID Number of Cells Class Name
4 951 water
# Layers 1 2 3
# Means
1.1504 0.0515 0.0873
# -----------------------------------------------------------------
# Class ID Number of Cells Class Name
5 969 agri_field
# Layers 1 2 3
# Means
32.4675 232.7781 85.4149