AUTOMER

Description

Parameters

Parameter descriptions

Separability Measures

Specifies the type of separability measure used. For more information, see the DETAILS section.

Input: Class Signature Layer: Input class signature segment

Specifies two to 256 signature segments for which separabilities are calculated.

All segments must be created using the same set of input channels. Duplicate segments are not allowed.

Separability Threshold

Optionally specifies a separability threshold value for the report. If this parameter is not specified, a matrix of separabilities for all class signature pairs is printed. If this parameter is specified, separabilities for only those class signature pairs with separabilities less than the defined values are printed (one pair per line).

Output: Merged Signature Layer(s): Output signature segment

Specifies the output signature segment(s) to receive the merged signatures.

Report

Specifies where to direct the generated report.

Available options are:

• LOG: generates a report on the terminal (default)
• <modulename.RPT>: appends a report to the module's RPT file
• DISK: generates a report on file "IMPRPT.LST"

Details

AUTOMER calculates the separability between each pair of classes among two to 256 classes and merges signature pairs that have the lowest separability among a set of signatures until a minimum separability threshold value is reached.

When the merging is complete, AUTOMER prints a brief report that shows the number of segments merged, the final signature segments, and the minimum separability.

Class signature segments can be created using KCLUS, or ISOCLUS. Pairs of classes with poor separabilities should be merged before performing a classification with MLC.

If a separability threshold (SEPTHRES) is specified, only those class signature pairs with a separability less than SEPTHRES are printed. This is useful if a large number of signature segments are specified, and you want to list only those signatures that should be merged.

The Transformed Divergence (TD) measure is a popular empirical measure that is computationally simpler than the Bhattacharyya Distance (BD) measure; however, the BD measure is more theoretically sound because it is directly related to the upper bound of the probabilities of classification errors.

Both TD and BD separability measures yield real values between 0 and 2, where 0 indicates complete overlap between the signatures of two classes and 2 indicates a complete separation between the two classes. TD and BD measures are monotonically related to classification accuracies. The larger the separability values, the better the final classification results (using MCL or MINDIS). The following rules are suggested for the possible ranges of separability of value 'x':

0.0 < x < 1.0   (very poor separability)
1.0 < x < 1.9   (poor separability)
1.9 < x < 2.0   (good separability)

Poor separability (1.0 < x < 1.9) indicates that the two signatures are separable to some extent. It is prefereable, however, to improve separability if possible. Low signature separability is usually caused by improper combinations of image bands or training sites, or both, which have large internal variability within each class.

Algorithm

To determine the separability between a pair of class signatures, Transformed Divergence (TD) and Bhattacharyya Distance (BD) are used. They are calculated from class means and covariance matrices as follows:

         TD Measure
  ______________________________

  TD(i,j) = 2*[1-exp(-D(i,j)/8)]

        BD Measure
  ____________________________

  BD(i,j) = 2*[1-exp(-a(i,j))]

The function calculates the separabilities of all the possible pairs among 'n' classes, where 'n' is the number of classes.

References

Richards, J. A., "Remote Sensing Digital Image Analysis". Berlin: Springer-Verlag, 1986.