We may specify multiple features to be extracted from each object/region. Say, we wish to extract two spectral indices that we have defined already and some shape representation of each object. First, we need to select a specific spectral index and then choose the desired representation in the Add representation combo to include it in the list.


Available feature types

  • Mean spectrum: mean spectrum computed using all pixels specified
  • Spectral feature histogram for the selected spectral feature, exists of a histogram of 20 equally sized bins between the min and max boundaries defined in the Visualization panel
  • Spectral feature mean for the selected spectral feature
  • Foreground pixel count
  • Class fraction for the class selected in the Classes list
  • Regression output for the selected regression variable
  • Object count within the region
  • Object shape: Feret diameter: shape representation providing minimum and maximum caliper distance for the object mask
  • Object shape moments: a set of 7 Hu shape moment invariants and an object eigenvalue ratio
  • Object shape circularity describes how far certain object shape is from a circle, three features are provided, namely Circularity, the Area/Perimeter ratio and the Perimeter


In our example, we defined the following features:



Below a screenshot of the exported Excel file with indicated 5 feature groups: