Sample Analysis and Image Classification

This page describes the window "Image Classification", used for sample analysis and image classification.

To make it easier, we will subdivide the procedures in three steps:


See also:


Sample Analysis:

  • after accessing the option Classification... (inside the "Classification" window), it is possible to analyze the samples before executing the classification itself;
  • type in a Name to be copied to the classified image and click on Create;
  • click on Sample Analysis... at this window. After a few seconds, the window "Sample Analysis" will be presented;
  • after analyzing the samples and deciding to classify the whole image you can perform the classification.


Pixel Based Classification

  • after having analyzed the samples you can execute the classification. The samples analysis is not mandatory!
  • choose the Classifier: Maxver, Maxver-ICM, Dist. Euclidiana or KMedias;
  • if Maxver - provide the Acceptance-Threshold (%): 100, 99.9, 99, 95, 90 or 75;
  • if Maxver (ICM) provide the Percentage of Change: 1 to 100;
  • if Euclidean Distance - analyze the samples you wish;
  • if Kmedias - provide the number of themes (2 to 55) and the number of iterations. If you want that the classifier is based only on an area of the image, choose Area of Acquisition;
  • Set the Name to be created for classified image and click Category... (If you want to change the category where the image is saved Classified);
  • click on Classify;
  • after classifying an image, you can see it from the "Control panel" and, if you wish, perform a post-classification

NOTE: The percentage of change is the maximum number of iterations during the classification process for changing the pixels values. For instance, for a value up to 30, there will be new iterations and thus new changes. The greater the number of interactions, the quicker the classification.

 

Image Classification


Classification by Region:

  • choose the Classifier: Isoseg, Battacharya , ClaTex, SRN file or Histograma.
    • if Isoseg - provide the Acceptance-Threshold (%): 100, 99.9, 99, 95, 90 or 75;
    • if Battacharya - provide the Acceptance-Threshold (%): 100, 99.9, 99, 95, 90 or 75; and analyze the samples you wish;
    • if ClaTex - define the texture file by clicking on Texture File...
    • if SRN file - click on Import SRN File...
    • if Histogram - define the number of themes ( 2 to 55 );
  • Set the Name to be created for classified image and click Category... (If you want to change the category where the image is saved Classified);
  • click on Classify;
  • after classifying an image, you can see it from the "Control panel" and, if you wish, perform a post-classification

Note: Defining SRN File ( Neural Network System )
This window allows to search the saved file (extension .res) using the Neural Network software.

Mapping to Thematic Classes