Per pixel supervised classification
From AWF-Wiki
(Difference between revisions)
(→Using QGIS and OTB processing toolbox) |
(→Training phase) |
||
| Line 7: | Line 7: | ||
* In the search engine of Processing Toolbox, type {{typed|text=TrainImages}} and open '''TrainImagesClassifer'''. | * In the search engine of Processing Toolbox, type {{typed|text=TrainImages}} and open '''TrainImagesClassifer'''. | ||
* In the {{button|text=Input Image List}} select one (or optional: several) multi-band raster (multispectral images). | * In the {{button|text=Input Image List}} select one (or optional: several) multi-band raster (multispectral images). | ||
| − | * In the {{button|text=Input Vector Data List}} | + | * In the {{button|text=Input Vector Data List}} select a file containing training area polygons in format GPKG or SHP. |
* If you do not have independent validation data: leave {{button|text=Validation Vector Data List}} empty. | * If you do not have independent validation data: leave {{button|text=Validation Vector Data List}} empty. | ||
* Bound sample number ba minimum: {{typed|text=1}}. The class with the minimum number of pixels determines the sample size of all other classes. Changing this value to {{typed|text=0}} does not have an effect (current bug in QGIS3.16). | * Bound sample number ba minimum: {{typed|text=1}}. The class with the minimum number of pixels determines the sample size of all other classes. Changing this value to {{typed|text=0}} does not have an effect (current bug in QGIS3.16). | ||
* Training and validation sample ratio: {{typed|text=0.25}}. This is a 4-fold cross-validation with split 0.75 per cent for training and 0.25 per cent for testing. | * Training and validation sample ratio: {{typed|text=0.25}}. This is a 4-fold cross-validation with split 0.75 per cent for training and 0.25 per cent for testing. | ||
| − | * Type {{typed|text= | + | * Type {{typed|text=Code}} in the {{button|text=Field containing the class integer label for supervision}} text field. |
* As Classifier to use for training choose Support Vector Machine {{button|text=libsvm}} from the drop down list. | * As Classifier to use for training choose Support Vector Machine {{button|text=libsvm}} from the drop down list. | ||
* The SVM Kernel Type is {{button|text=Linear}}. | * The SVM Kernel Type is {{button|text=Linear}}. | ||
Revision as of 13:31, 18 June 2021
Contents |
Using QGIS and OTB processing toolbox
- Using OTB version > 7.2.0, input vector file format of training data can be GeoPackage GPKG or ESRI Shapefile.
- The column containing class label needs to be an integer (whole number).
Training phase
- In the search engine of Processing Toolbox, type TrainImages and open TrainImagesClassifer.
- In the Input Image List select one (or optional: several) multi-band raster (multispectral images).
- In the Input Vector Data List select a file containing training area polygons in format GPKG or SHP.
- If you do not have independent validation data: leave Validation Vector Data List empty.
- Bound sample number ba minimum: 1. The class with the minimum number of pixels determines the sample size of all other classes. Changing this value to 0 does not have an effect (current bug in QGIS3.16).
- Training and validation sample ratio: 0.25. This is a 4-fold cross-validation with split 0.75 per cent for training and 0.25 per cent for testing.
- Type Code in the Field containing the class integer label for supervision text field.
- As Classifier to use for training choose Support Vector Machine libsvm from the drop down list.
- The SVM Kernel Type is Linear.
- SVM Model Type is csvc
- Switch checkbox Parameters optimization off. The optimization results in a higher accuracy but takes much time (> 1 hour computation).
- In the Output model specify a model file: e.g. svm.model
- Click Run.
- Click on the Log tab and inspect the model quality measures: Precision, Recall, F-score and Kappa index.
Classification phase
- In the search engine of Processing Toolbox, type ImageClassifier and double click ImageClassifier.
- Set an multiband image as Input image.
- Set Input _mask to blank (top of drop-down list).
- Set svm.model as Model file.
- Set the number of classes in your model: this is the number of unique classes in your training vector file.
- Set Output pixel type to uint8
- Save the Output image as svm_classification.tif.
- Uncheck Confidence map: Open output file after running algorithm.
- Uncheck Probability map: Open output file after running algorithm. Run.
- Find the output svm_classification.tif in the QGIS map canvas.
- Right click on the layer svm_classification and select Properties --> Symbology --> Style --> Load Style.
- Select the style file classified.qml. OK.
Postprocessing
This application uses a majority filter with a circular structure element.
- In the search engine of Processing Toolbox, type regularization and double click ClassificationMapRegularization.
- Define Input classification image as an classification output raster.
- Set Structuring element radius to 2 pixels.
- Set Output pixel type to uint8.
- Save the Output regularizd image as svm_classification_majority.tif.
- Click Run.
- Find the output svm_classification_majority.tif in the QGIS map canvas.
- Right click on the layer svm_classification_majority in the and select Properties --> Symbology --> Style --> Load Style.
- Select the style file classified.qml. OK.
Using OTB standalone
Training phase
- Navigate in the Windows explorer to your Folder OTB-7.2.0-Win64. Double click the Windows-Batchdatei mapla to open Monteverdi Application Launcher.
- In the search engine of mapla, type TrainImages and double click TrainImagesClassifer.
- In the Input Image List click on + and select a (or optional: several) multispectral images: Subset_S2A_MSIL2A_20170619T_MUL.tif .
- In the Input Vector Data List choose a vector polygon file with training areas in file format GPKG or SHP.
- In the Output model specify an output model file: e.g. lucc_rf.model
- In the Bound sample number by minimum field type 0.
- Set the training and validation sample ratio to 0.25.
- Mark C_ID in the Field containing the class integer label (C_ID refers to the column that contains the LUC code in the training and validation vector file).
- Choose Shark Random forests classifier from the drop down list as Classifier to use for the training.
- Check user defined Random seed and enter any positive integer value. This initializes a pseudorandom number generator which may be used to reproduce results.
- Click on Execute.
- Click on the Log tab and inspect the model quality measures: Precision, Recall, F-score and Kappa index.
Classification phase
- In the search engine of mapla, type ImageClassifier and double click ImageClassifier
- Set Subset_S2A_MSIL2A_20170619T_MUL.tif as Input image.
- Set lucc_rf.model as Model file.
- Save the Output image as rf_classification.tif.
- Adjust the Number of classes in the model to the number of unique classes in the training vector file.
- Add rf_classification.tif to QGIS canvas.
- Download the style file classified.qml from Stud.IP.
- Right click on the layer rf_classification and select Properties --> Style --> Style --> Load Style.
- Select the style file classified.qml. OK.
Postprocessing
This application uses a majority with a circular structure element.
- In the search engine of mapla, type, regular and double click ClassifcationMapRegularization.
- As theInput classification image define the output raster file of the classification phase.
- Save the Output regularizd image as rf_classification_majority.tif.
- Set Structuring element radius to 2 pixels.
- Set Output pixel type to uint8.
- Click Execute.
- Add rf_classification_majority.tif to QGIS canvas.
- Download the style file classified.qml from Stud.IP.
- Right click on the layer rf_classification_majority and select Properties --> Style --> Style --> Load Style.
- Select the style file classified.qml. OK.
