By export visualizations as float images, we get Matlab .mat files extracted from our spectral images. The use-case is to define one or more custom feature indices, export the floating point data together with precise pixel labeling and perform further analysis or training of external models in Matlab, Python, or other machine learning environments.



In the following example, we defined three spectral indices. We selected the scans we wish to process and clicked File / Export / Export visualization (float image). We may then select a destination directory. For each selected image a .mat Matlab binary file is created with floating point spectral index content and separate pixel labels.



Example on Matlab side:


>> ls


.                     ..                    natural_objects1.mat  natural_objects2.mat 


>> load natural_objects1.mat

>> whos

  Name        Size                 Bytes  Class     Attributes


  cube      640x503x3            3863040  single              

  lab       640x503               321920  uint8       


>> figure; imagesc(cube(:,:,3)')

>> figure; imagesc(lab')



Comments:

  • For each exported image a .mat file will be present in the destination directory.
  • Each of these files contain a cube and lab variable.
  • The cube variable contains a band (3rd dimension) for each of the spectral indices.
  • Note that we transpose the image content using ' operator to visualize images in the same way as in perClass Mira.
  • The lab variable contains per-pixel labels defined in perClass Mira, class indices may directly map to the class list in perClass Mira. Zero is the "unknown" class such labels are not present.