Multi Source Classifier
Allows the user to perform radar raster data classification using following the methodology described by:

B. C. Braga, C. d. C. Freitas and S. J. S. Sant'Anna, "Multisource classification based on uncertainty maps," 2015 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Milan, Italy, 2015, pp. 1630-1633, doi: 10.1109/IGARSS.2015.7326097.

It is accessible through:
   Processing → Radar Raster Processing → Multi Source Classifier...

  1. Number of images: Set this box value with the correct number of input rasters used in the process.
  2. Segmented Image: Click on the button "..." to select the input segmented raster.
  3. Training Samples: Click on the button "..." to select the input training samples text file.
  4. Test Samples: Click on the button "..." to select the input test samples text file.
  5. Combination type: How the build the multi-source classification procedure (sum, multiplication,minimum,Hellinger,fuzzy).
  6. Confidence level: The significance level.
  7. Horizontal lag:  The horizontal correlation value.
  8. Vertical lag: The vertical correlation value.
  9. Repository: Click on the button "..." to select the output classified segmented raster location.
  10. Save on Disk: Enable/disable the creation of extra output data files.
  11. Click on the "OK" button to go the the next window.
Te next window allows the selection of options related to each input raster data file:
  1. Click on the button "Input image" to select the current input data raster.
  2. Statistic model: The used input raster data model (Gaussian,Gamma,IntensityPair,Wishart).
  3. Stochastic Distance: The stochastic distance type.
  4. Parameters:
    1. ENL:  Equivalent Number of Looks.
    2. Beta: Beta values for Renyi distance.
    3. Covariance Matrix Order: An integer number for the lexicographic vector and covariance matrix order.
  5. Click on the "OK" button to start the operation.