Classify

Dialog box

The classify operation performs a multi-spectral image classification according to training pixels in a sample set. The following classification methods can be used: Box classifier, Minimum distance, Minimum Mahalanobis distance, Maximum Likelihood, and Maximum Likelihood with Prior Probabilities.

Dialog box options:

Sample set:

Select an input sample set. Open the list box and select the desired sample set, or drag a sample set directly from the Catalog into this box.

Classification method:

Select a classification method: Box Classifier, Minimum Distance, Minimum Mahalanobis Distance, Maximum Likelihood or Prior Probability. For more information, see Classify : functionality.

Multiplication factor:

For the Box Classifier method, type a multiplication factor, usually larger than 1.

The multiplication factor allows you to make boxes drawn around class means on the basis of their standard deviations a bit wider.

Threshold distance:

For the Minimum Distance, Minimum Mahalanobis distance, Maximum Likelihood and Prior Probability classification methods:

  • Select this check box and subsequently type a threshold distance when pixels with a spectral signature that is not similar to any of the training classes, should not be classified.
    • When, for a pixel to be classified, the spectral distance towards a class mean:

    • is smaller than or equal to the threshold, the pixel is acceptable for that class; the pixel will be classified as that class when there is no other class at an even smaller spectral distance;
    • is larger than the threshold, the pixel is rejected and will not classified as that class.
  • Clear this check box when all pixels should be classified.

 

Tips:

  • The larger you choose the threshold value, the more pixels will be classified.
  • The calculation of spectral distances towards class means depends on the selected classification method.
  • For an example of the meaning of the threshold value for the Minimum Distance classifier, see Classify : algorithm.

Table:

For the Prior Probability classification method: select the table which contains the column with the prior probability values.

The table must use the same class domain as the sample set.

Column:

For the Prior Probability classification method: select the column with the prior probability values.

Ideally, the sum of probabilities equals 1. When this is not the case, the program will internally normalize the probability values.

Output raster map:

Type a name for the output raster map that will contain the outcome of the classification.

Description:

Optionally, type a description for the output map. The description will appear in the status bar of the Main window when moving the mouse pointer over the map in a Catalog, and in the title bar of a map window when the output map is displayed. If no description is supplied, the output map will use its own definition as description.

When you click the Show button, the dependent output map will be defined, calculated and shown. When you click the Define button, the dependent output map will only be defined; if necessary the map will be calculated later, for instance when the map is opened to be displayed.

Tip:

To assess the accuracy of the classification, you can prepare a confusion matrix. For more information, see How to calculate a confusion matrix.

See also: