Exercise: Classify an image in QGIS

For issues, bugs, proposals or remarks, visit the issue tracker.

Tutorial Data Set

We will use several images:

  • WINDOWED_SPOT_XI_VIETNAM_011.tif, etc.. for predicting
  • reference.tif for training.

We have several tif images at our disposal: band 11, 12, 13 and 14 all saved as separate bands, we have an image with band 11 and 12, and an image with all 4 bands.

The classification in the reference.tif file is the following:

  • 1: Water
  • 2: Mangrove/aquaculture
  • 3: Urban/bare white ground
  • 4: Agriculture
  • 5: Forest
  • 6: Mountaintops/mining
  • 0: No data

Goal

We will do a supervised classification MLP Classifier.

_images/image_bands.PNG

Hint

Use the style file in the exercise folder (classification_style_raster.qml) for easy styling.

Run the Neural Network MLPClassifier in QGIS

We need the following input for the GUI:

  • The image for classification: make sure you select either the image with 4 bands, or the 4 separate bands, … but do not select them twice!
  • The raster for training: reference.tif
  • Number of iterations: 2000
  • No-data-value: 0 (do not forget to set this one!)
_images/image_gui.PNG

After the training, the network error is shown as a plot like:

_images/image_network_error.PNG

The resulting image should look something like this:

_images/classified_image.PNG

You will also get an output on the log screen with the classification kappa

Average accuracy: 0.9804091040046097
Kappa class 1: 0.999608231804873
Kappa class 2: 0.998816901840517
Kappa class 3: 0.943025695066823
Kappa class 4: 0.9422660833150642
Kappa class 5: 0.9909993784914707
Kappa class 6: 0.8952379269755105
Average Kappa: 0.961659036249043

              precision    recall  f1-score   support

           0       1.00      1.00      1.00      1685
           1       1.00      1.00      1.00       985
           2       0.96      0.93      0.95       584
           3       0.93      0.98      0.95      1121
           4       0.99      0.99      0.99      2351
           5       0.97      0.84      0.90       216

   micro avg       0.98      0.98      0.98      6942
   macro avg       0.98      0.96      0.97      6942
weighted avg       0.98      0.98      0.98      6942
 samples avg       0.98      0.98      0.98      6942